The Phillips Curve: a Relation between Real Exchange Rate ......Phillips Curve Period $ deval...

The Phillips Curve: a Relation between Real ExchangeRate Growth and Unemployment∗

François Geerolf†

UCLA

August 1, 2019Last Version

AbstractThe negative relationship between inflation and unemployment (also known as the Phillips curve)

has been repeatedly challenged in the last decades: missing inflation in 2013-2019, missing deflation in2007-2010, missing inflation in the late 1990s, stagflation in the 1970s, contrasting with always strongregional Phillips Curves. Using data from multiple sources, this paper helps to solve many empiricalpuzzles by distinguishing between fixed and flexible exchange rate regimes: in fixed exchange rateregimes, inflation is negatively correlated with unemployment but this relationship does not hold inflexible regimes. By contrast, there is a negative correlation between real exchange rate growth andunemployment, which remains consistent in both fixed and flexible regimes. These crucial observationshave important implications for identifying the source of business cycle fluctuations and for normativeanalysis.

Keywords: Phillips curve, unemployment, inflation.JEL classification: E2, E24, E3, E31, F3.

IntroductionThe Phillips Curve is named after A.W. Phillips who first documented a negative correlation between inflationand unemployment in the United Kingdom (Phillips (1958)).1 It is a pillar of the neoclassical synthesis,according to which the Phillips Curve traces a menu of short-run options between inflation and unemployment,an aggregate supply curve. By increasing aggregate demand through monetary or fiscal policy, policymakerscan boost employment for some time, at the cost of higher inflation. The Phillips Curve trade-off betweeninflation and unemployment is taught in most undergraduate textbooks that include some treatment ofKeynesian economics (for example, Mankiw (2015), Blanchard (2016a), Jones (2017)), and the New-KeynesianPhillips curve is a reference point for modern Dynamic Stochastic General Equilibrium (DSGE) modelswith sticky prices (Smets and Wouters (2007)), at the heart of most central banking policy and research.For example, the Phillips curve is used to interpret the source of business cycle fluctuations: aggregatedemand shocks are supposed to be inflationary, and reduce unemployment. The “missing deflation” duringthe financial crisis of 2007-2009 has led some economists to question whether the financial crisis shouldprimarily be interpreted as coming from a shortfall in aggregate demand. (Beraja, Hurst, and Ospina (2016))

Yet despite its impressive impact and widespread adoption, the empirical relevance of the Phillips curvehas been challenged at numerous occasions, and the Phillips curve is subject to repeating controversies.Some of these controversies are central to the history of macroeconomic thought: for example, the 1960s-70scontroversy about stagflation led to the adoption of the expectations-augmented Phillips curve and the notionof a so-called “natural rate of unemployment.” Moreover, the Phillips curve is viewed as a test for Keynesianeconomics: in the 1970s, the coexistence of high unemployment and high inflation led Robert Lucas and

∗I thank Andy Atkeson, Martin Beraja, Yannick Kalantzis, Pierre Jacquet, Eric Monnet, Andy Neumayer, Martin Uribe,Melanie Wasserman, and seminar participants at UC Riverside, Universidad Torcuato Di Tella, and the SED 2018 MexicoMeetings for useful comments. This is preliminary and incomplete, please read the most recent version of the draft here.

†Contact: [email protected] U.S. Phillips curve is usually credited to Samuelson and Solow (1960). In fact, Fisher (1926) documented the U.S.

correlation between inflation and unemployment well before Samuelson and Solow, and even before A.W. Phillips.

1

https://fgeerolf.com/phillips.pdf

https://fgeerolf.com/phillips.pdf

mailto:[email protected]

Thomas Sargent to ask what would come “after Keynesian macroeconomics” (Lucas and Sargent (1979)).In the last few years, there has been a new controversy about the Phillips curve, which is still ongoing.It has even been discussed in Congress by Representative Alexandria Ocasio-Cortez, with Fed ChairmanJerome Powell acknowledging that the Phillips curve is now merely a “faint hearbeat”, and discussingimplications for monetary policy.2 Indeed, inflation has not increased despite unprecedented fiscal stimulusand low unemployment, so that the Phillips curve seems absent at the aggregate level. Looking back onthe performance of macroeconomics since the financial crisis, Paul Krugman (2018) has argued that thereexists “a big failure in our understanding of price dynamics.” At the same time, some scholars have notedthat regional Phillips curve have remained strong (McLeay and Tenreyro (2019), Hooper, Mishkin, and Sufi(2019)).

In this paper, I argue that underlying the Phillips curve is in fact a negative correlation between real exchangerate growth and unemployment. My hypothesis allows to reconcile these apparently conflicting observationsabout the Phillips curve. Under fixed exchange rates, real exchange rate growth equals inflation, whichimplies the negative correlation between inflation and unemployment found by A.W. Phillips in the U.K.under the Gold Standard, and by Samuelson and Solow in the U.S. under Bretton Woods. I show thatmore generally, Phillips curves always exist and are strong in fixed exchange rate regimes, such as acrossEuropean countries in the Euro area. Regional Phillips curves, for example across U.S. states or MetropolitanStatistical Areas, are also consistent with a negative correlation between real exchange rate growth andunemployment, as in this case too, regional inflation corresponds to a change in relative prices across regions.That the change in relative prices comes entirely from non-traded goods such as house prices and rents, whichoverwelmingly explain relative inflation rates across cities and states, both in nominal wages and prices, iswell known and documented in the urban and economic geography literature (Moretti (2013)). Under flexibleexchange rates however, there need not exist a correlation between overall price inflation and unemployment,because nominal exchange rates make the nominal price of traded goods fluctuate in the local currency. Idocument that there does always exist a robust relationship between the relative price of non-traded goodsand that of traded goods (rents and housing) and unemployment, as well as between real exchange rates andunemployment. However, this relationship between relative prices is mediated by the nominal exchange rate.It is on average offset by a positive relationship between traded goods price inflation and unemployment,coming from movements in the nominal exchange rate: on average, nominal exchange rates depreciate whenunemployment is high, which goes against the increase in the relative price of housing. In other words, thePhillips curve corresponds to a relationship between relative prices and unemployment, not to a monetaryphenomenon.

For concreteness, I now take the example of the missing U.S. inflation since 2013, which Janet Yellen wasalready calling in 2017 “the biggest surprise in the U.S. economy” (Yellen (2017)). Despite an unprecedentedfiscal stimulus in an economy considered above potential, U.S. inflation has not risen above 2%, which hasput the Federal Reserve in a difficult position about whether it should raise rates or not. As a consequenceof that stimulus, the U.S. dollar has appreciated in nominal terms against many of the currencies of itstrading partners, making imported goods cheaper: the price of traded goods in dollars has fallen relatively.In contrast, rents and house prices have indeed gone up in dollar terms. Overall CPI inflation has thus notrisen by as much: the increase in rent prices has contributed to inflation, while falling traded goods priceshave contributed to deflate the economy. This paper shows that this pattern is in fact general, and thatthis is what the Phillips curve is ultimately about. With fixed exchange rates, or within regions of the sameeconomy, traded goods prices are approximately constant so that the relative increase in the price of housingresults in an increase of overall CPI inflation. In particular, this explains Phillips’ original correlation. Underflexible exchange rates, the dollar price of traded goods can move through the nominal exchange rate, sothat overall inflation in dollars can correlate to unemployment or not, depending on nominal exchange ratemovements. I in fact show that most, if not all, Phillips curve related controversies can in fact be very simplyunderstood using this fixed versus flexible exchange rate dichotomy, and distinguishing between traded andnon-traded goods. This finding has important implications for interpreting business cycles: using a Phillipscurve framework, Beraja, Hurst, and Ospina (2016) argue that the missing deflation during the financialcrisis implies that an adverse supply shock has taken place at the aggregate level, since cross-regional Phillips

2https://www.youtube.com/watch?v=hyCsxzd2a50.

2

https://www.youtube.com/watch?v=hyCsxzd2a50

curve were strong. According to the real exchange rate Phillips curve, the 2007-2009 could have been a pureaggregate demand shock, even if it did not result in any deflation. Similarly, Zidar (2019) interprets taxmultipliers as arising from aggregate supply effects, whereas the rise in wage inflation, house price inflation,and constancy of the real wage is in fact typical of the response to aggregate demand shocks: therefore, hisresults of large tax multipliers following tax cuts to the bottom 90% of workers can be interpreted as arisingfrom disposable income effects.

In support of my hypothesis, I first document the strength of Phillips curves across countries, depending ontheir exchange rate regimes. I show that the relation between inflation and unemployment exists in fixedexchange rate regimes more broadly in a panel of 35 OECD economies, while it does not on average in flexibleexchange rate regimes. Unlike traditional Phillips curves, which have become ever noisier over the years, andare very sensitive to specifications, this relationship is very strong and robust, and holds regardless of whetherone examines headline, core, harmonized inflation rates, and consumption or GDP deflators. Moreover, Ishow that this fact is mostly driven by the price of tradables, which fail to have a Phillips curve correlationin flexible exchange rate regimes. In contrast, the price of non-tradables (such as the price of housing, rentsand services) is everywhere correlated with unemployment, again supporting the main finding of the paperthat Phillips curves really are about relative prices, and a correlation between the relative price of non-tradedgoods versus traded goods and unemployment. I next consider event studies and large shocks, which arecloser to “natural experiments”: in particular, the missing deflation in 1933, or the 1970s stagflation areexamples of a change of exchange rate regimes, associated with a shift in the Phillips curve. I also show thatthe response of the economy to “large shocks”, and the associated failure of the price Phillips curve, such asthe 2007-2009 missing deflation, and the 2013-2019 missing inflation, are also supportive of the model. I thenuse monetary and fiscal policy shocks in the United States to investigate the source of the Phillips curve, andtrace it again mostly to the relative price of non-traded goods. I then move on to regional Phillips curve.I show, in line with a growing literature (most recently, Hooper, Mishkin, and Sufi (2019)), that regionalPhillips curve are very strong in the United States, as well as across European countries. However, unlike theexisting literature, I relate these regional Phillips curve to the exchange rate regime: this paper shows thataggregate, flexible exchange rate relations between inflation and unemployment are completely different fromregional ones, as they are mediated by the nominal exchange rate.

This finding of a real exchange rate Phillips curve has many implications, of which I will only list four. First,the real exchange rate Phillips curve implies that the economy can in fact be thought of as having highaggregate demand even when overall CPI inflation is low, as the relative price of housing goes up, such as atthe end of the 1990s (the “missing inflation” under Alan Greenspan) and the housing boom of the 2000s. Thereal exchange rate Phillips curve implies that an appreciating dollar implies that aggregate demand shocksneed not be inflationary, and that the Federal Reserve should monitor at the relative price of non-tradedgoods versus traded goods, if the objective is aggregate demand stabilization. Similarly, one cannot inferfrom the “missing deflation” during the 2007-2009 that the U.S. simultaneously experienced an aggregatedemand and an aggregate supply shock, but simply that all countries experienced a similarly sized negativeaggregate demand shock, and that the U.S. has a flexible exchange rate. Diagnozing the cause of businesscycle fluctuations is not just of academic interest, but it is very important for designing an appropriate policyresponse: during the financial crisis, the “missing deflation” led some economists to argue against the use ofaggressive fiscal and monetary policy. Second, the associated costs of high aggregate demand can be in anoversized real estate, construction, and non-traded sector, rather than the usual channel of price dispersion ofthe New-Keynesian model. In a context of dynamic inefficiency (Geerolf (2013b)) or demand-side secularstagnation however (Hansen (1939), Summers (2013), Geerolf (2019)), the problem lies more in low aggregatedemand in surplus countries than in too high aggregate demand in deficit countries: there is a global excess ofsaving over investment, and surplus countries simply “export their way out” of this problem, by finding storesof value abroad. In fact, these issues of surplus countries’ adjustment was central to John Maynard Keynes’thinking around the creation of the International Monetary Fund, and part of the so-called Keynes Plan atBretton Woods. Third, the real exchange rate Phillips curve highlights a trade-off between unemploymentand real exchange rates which is of primary concern to policymakers, and is often behind the reluctanceto stimulate the economy, even when inflation is low, as they worry it might hurt “competitiveness.” Thereal exchange rate Phillips curve highlights that one negative consequence of aggregate demand stimulatingpolicies sometimes is trade deficits, loss of competitiveness, and an oversized non-traded sector. However,

3

again, I shall discuss that in a context of global secular stagnation, this problem is merely a coordinationproblem, which comes from surplus countries’ deficient aggregate demand. Fourth, this new interpretation ofthe Phillips curve implies that the stagflation experienced by the U.S. in the 1970s was no failure of Keynesianeconomics, nor a triumph of rational expectations and the Lucas (1976) critique, but simply coming from thedepreciation of the dollar allowed by the exit from the Bretton Woods system of fixed exchange rates.

The rest of the paper proceeds as follows. Section 1 provides an array of evidence using correlations, focusingon the importance of the exchange rate regime for the existence of different types of Phillips curve. Section 2presents within-country event studies corroborating cross-country evidence. Section 3 shows some evidencerelating to large events, where other sources of variations are swamped. Section 4 shows that real exchangerates Phillips curve also appear conditionally as a result of monetary and fiscal aggregate demand shocks.Section 5 discusses new and existing evidence concerning regional data, in majority in the United States. InSection 6, I show that the real exchange rate Phillips curve has important consequences for positive analysis,and in particular for identifying the source of business cycle fluctuations. I also discuss the implications ofthis empirical finding for normative analysis: the real exchange rate Phillips curve shows that the costs ofhigh aggregate demand are in the form of too high a real exchange rate, real estate overvaluations, and anoversized non-traded sector. However, assuming secular stagnation (a global excess of saving over investmentdemand), the burden of the adjustment should fall on surplus not deficit countries.

1 Phillips Curve CorrelationsIn this section, I present Phillips curve correlations, in the spirit of early work by Fisher (1926), Phillips (1958)and Samuelson and Solow (1960), implicitely assuming that most if not all shocks are aggregate demandshocks.3 I first look at price and wage Phillips curve in different types of exchange rate regimes. A robustfinding is that price and wage Phillips curve are a strong feature of fixed exchange rate regimes, but that theyare insignificant, sometimes positive, in any case always much noisier with flexible exchange rates. I then focuson the United States and the United Kingdom, given that early Phillips curves were documented in thesetwo countries, and that many discussions around the Phillips curve revolve around the U.S. macroeconomichistory. Finally, I show that the relative price of non-traded goods (mostly house prices and rents) are verystrongly and robustly negatively correlated to unemployment across exchange rate regimes.

1.1 Price and Wage Phillips Curve in Different Exchange Rate RegimesIn this section, I use the OECD’s Consumer Price Indices, Economic Outlook, Quarterly National Accounts,Main Economic Indicators from 35 countries, and merge this cross-country database to data on exchangerate regimes constructed by Ilzetzki, Reinhart, and Rogoff (2019). According to their classification, exchangerate regimes are divided between “Fixed / Peg”, “Crawling Peg”, “Crawling Band”, and “Floating”. Table 49in appendix F.1 gives a detail of exchange rate arrangements corresponding to this coarse classification.

Original Price Phillips Curves. The first suggestive piece of evidence in favor of the main thesis in thispaper is exposed in Table ?? and Table ??. These two tables present results from the original Phillips Curveregression, with inflation being 1-Year inflation πit, and Uit being the unemployment rate, with country-levelfixed effects δi:

πit = α+ δi + βUit + εit.

where the regression coefficient β is the slope of the Phillips Curve (PC Slope in the tables).

Table 2 shows the results from these Phillips curve regressions, run separately for each type of exchange rateregime. Because of fixed effects, the Phillips curve is here identified from within country variation. Withfixed effects, the Phillips Curve appears to be present for the “Fixed / Peg” type of regime, as well as the“Crawling Band”, but not for floating exchange rates, and crawling pegs. The explained variance of inflationexplained by unemployment (the adjusted R2) is greater for fixed exchange rates.

3For readers whose prior belief is that business cycles are mostly driven by technology shocks, correlations conditional onidentified aggregate demand shocks will be computed in section 4.

4

Table 1: Original Price Phillips Curve: Headline inflation and Unemployment, No FixedEffects

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.26*** -8.4 11.3% 547Crawling Peg -0.21*** -2.7 1.8% 346Crawling Band -0.04 -0.5 -0.2% 451

Floating 0.28*** 2.7 4.7% 129

Table 2: Original Price Phillips Curve: GDP Deflator Growth and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.35*** -8.4 11.4% 547Crawling Peg -0.46*** -4.0 4.1% 346Crawling Band -0.75*** -4.4 3.9% 451

Floating -0.31** -2.1 2.5% 129

Table ?? runs the same regression, but now without country-fixed effects:

πit = α+ βUit + εit.

Therefore, the identification does not only rely from within country variation, but also from across countryvariation. This reveals an even stronger picture: under fixed exchange rates, we do observe a significant Phillipscurve: when the unemployment rate rises by 1%, inflation falls by −0.13% on average. This relationshipdecreases continuously when one moves from fixed exchange rates to floating exchange rates. In fact, whenone looks at floating exchange rates, the relationship between inflation and unemployment even turns positive,which is opposite to the Phillips curve. The positive relationship between inflation and unemployment issignificant and unemployment “explains” a large fraction of the variance in inflation, both across and withincountries.

Table 3: Alternative Inflation Definitions and Unemployment, by Exchange Rate RegimeInflation Concept Fixed Cr. Peg Cr. Band Float

Consumption Deflator Growth -0.22*** -0.19*** -0.06 0.3***Core Inflation -0.97** -0.47*** -0.08 0.32***

GDP Deflator Growth -0.26*** -0.21*** -0.04 0.28***Harmonized Core Inflation -0.1*** -0.29** 0.37** 0.07

Harmonized Headline Inflation -0.09*** -0.12 0.15 0.26Headline Inflation -0.24 -0.4*** -0.02 0.35***

Original Wage Phillips Curves. The original Phillips curve plotted wage not price inflation againstunemployment. In Table 4, I therefore look at the wage Phillips curve across exchange rate regimes. Onceagain, it can be seen that the correlation between nominal wage inflation and unemployment is very strong,and has a high R2 in fixed exchange rate regimes, not in floating exchange rate regimes. The magnitude ofthe correlation is also weaker in floating exchange rate regimes than in fixed.

Accelerationist (Expectations Augmented) Price Phillips Curves. Since the 1970s, following Phelps(1967) and Friedman (1968)‘s seminal contributions, and the 1970s stagflation, it is usually assumed that theredoes exist a permanent tradeoff between inflation and unemployment as postulated by the simple Phillipscurve, but that there only exists a temporary tradeoff. In other words, increasing unemployment does notimply that inflation falls, but simply that inflation decelerates: the reason is that agents’ expectations quicklyadapt to the new level of inflation. Therefore, I now test this accelerationist version of the Phillips Curve:

∆πit = βUit + εit

5

Table 4: Original Wage Phillips Curve: Wage Rate Inflation and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -0.32* -1.9 2% 128

Table ?? presents the results from such a regression. The accelerationist Phillips Curve does not appear verystrong in any of the exchange rate regimes. As previously, it is more significant in fixed change rate regimes,than in floating ones. The rest of the paper focuses on the original Phillips curves. In the United States too,the original Phillips curve appears stronger after the financial crisis than the 1970s accelerationist version, asargued by Blanchard (2016b).

Table 5: Accelerationist Price Phillips Curve: GDP Deflator Growth and Unemployment,No Fixed Effects

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.03 -1.5 0.2% 528Crawling Peg -0.08* -1.7 0.6% 336Crawling Band -0.04 -0.7 -0.1% 443

Floating -0.09 -1.5 0.9% 128

Accelerationist Wage Phillips Curves. Since the 1970s, the correlation between price inflation, or thechange in price inflation and unemployment is usually plotted. Table 6 shows that Phillips’ original correlationdoes not hold up anywhere, even in an expectations-augmented form. This again, strengthens my choice tostudy the original Phillips curve, rather than its accelerationist version, in the remainder of the paper.

Table 6: Accelerationist Wage Phillips Curve: Change in Wage Rate Inflation and Unem-ployment

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.09 -1.2 0.1% 277Crawling Peg -0.03 -0.3 -0.5% 177Crawling Band -0.25 -1.5 1.1% 96

Floating 0.05 0.4 -0.7% 127

1.2 House Price and Real Exchange Rate Phillips CurvesHouse Price Phillips Curves. While inflation and unemployment are not correlated in floating exchangerate regimes, some components of the Consumer Price Index still are. One such important example concernshouse prices, and other components of the CPI related to lodging. Table 7 shows indeed that in all fourtypes of exchange rate regimes, house prices are strongly negatively correlated to unemployment rates. Theappendix shows that this relationship is actually very strong and robust: it is even stronger when looking at2-Year inflation rates, or at measures of house prices other than the BIS’s.

Other CPI components. The OECD also provides some data on more disaggregated components of theConsumer Price Index. Table 8 presents the results from running the regression at the “sector” level:

πist = α+ δi + βsUit + εit,

where πist is sectoral inflation in sector s. I obtain one Phillips curve slope βs for every type of product s. Itcan be seen that the above insights generalize. Relatively local components of the price index are correlated

6

Table 7: House Prices (BIS) and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -1.38*** -3.6 8.7% 128

to the unemployment rate across exchange rate regimes, just as house prices or rents are. On the other hand,less local components are not correlated to the unemployment rate in flexible exchange rate regimes. Forexample

Table 8: 2-Year Inflation and Unemployment, by Exchange Rate RegimeCPI Concept Fixed Cr. Peg Cr. Band Float

CPI: All items non-food non-energy -0.58*** -0.91*** -1.03*** 0.1CPI: Energy -0.12 -0.41 -0.79** 0CPI: Goods -0.35*** 0.36 0 -0.9*CPI: Housing -0.73*** -1.21*** -0.72** -0.88***

CPI: Housing excl. imp. rent -1.07*** 0.05 -1.5*** -0.99***CPI: Services -0.69*** -0.61* 0.58** -1.03***

CPI: Services less housing -0.64*** -0.09 -0.63* -0.99***CPI: Services less housing (Housing excl. imp. rent) -0.52*** -0.99*** -4.02*** 0.17

Table 43 in the appendix presents more results at a more disaggregated level. Again, the results from theseregressions suggest a quite general pattern. The correlation between real exchange rates and consumptionbooms is also well-known in the literature, in the form of the Backus and Smith (1993) correlation, which inthe real business cycles literature is usually interpreted as evidence for a lack of risk-sharing.

1.3 United StatesThe usual account of the history of the Phillips curve says that although A.W. Phillips had documented thePhillips curve for the United Kingdom, Samuelson and Solow (1960) were the first to document it in theUnited States (even though as I argue in the conclusion based on Sleeman (2011), A.W. Phillips was probablynot happy with the Phillips curve). In fact, this account is not correct, at least because Fisher (1926) is anearlier study of the relationship between unemployment and inflation. Fisher (1926) even interpreted as acausal relationship: “The fact that deflation causes unemployment has been well recognised for many years inisolated instances, such as the great deflation of 1921 in America or the corresponding post-war deflation inGreat Britain, Czechoslovakia, or Norway. It has likewise been recognised that inflation carries with it agreat stimulation to trade and an increase in employment (or decrease in unemployment).” This correlationbetween prices and unemployment was, in fact, very well known at the time as argued by Robinson (1974):“In those days (unlike now) the leading symptom of a recession was a fall in prices.”

Table 9: U.S. Price Phillips CurvesPeriod Exchange Rate Regime U.S. PC Slope t-stat Adj. R2

1891-1933 Gold Standard -0.67*** -4.5 31.6%1933-1945 1933 Devaluation, War -0.26* -2.1 23.4%1945-1971 Bretton Woods -1.26** -2.1 12.2%1971-2016 Flexible Exchange Rates 0.28 1.1 0.4%

Figure 1 shows the U.S. Price Phillips Curve from 1891 to 1945, with data from Global Financial Data. At

7

the time, the U.S. was on the Gold Standard, with a fixed value of the dollar in terms of Gold, apart fromthe devaluation of the dollar in 1933 by Roosevelt. In fact, not only did Fisher (1926) “discover” the Phillipscurve much earlier than Phillips, and showed this correlation. He even related inflation or deflation to thepurchasing power of the dollar explicitely: “In short, facts and theory both indicate that in the dance ofthe dollar we have the key, or at any rate a very important key, to the major fluctuations in employment.If this conclusion be sound, we have in our power, as a means of substantially preventing unemployment,the stabilisation of the purchasing power of the dollar, pound, franc, lira, mark, crown, and any othermonetary units.” In doing so, Irving Fisher was actually stating a correlation between real exchange rateand unemployment. However, whether the relationship is causal as he stated or just simulateneous, we shalldiscuss later.

Gold Standard

Phillips Curve Period Missing Deflation

$ deval (Roosevelt)

−14%−12%−10%−8%−6%−4%−2%

0%2%4%6%8%

10%12%14%16%18%20%22%24%26%

1890 1895 1900 1905 1910 1915 1920 1925 1930 1935 1940 1945

Une

mpl

oym

ent R

ate,

Infla

tion

Inflation (Price)Unemployment Rate

Figure 1: U.S. Price Phillips Curve (1891-1945)

Figure 2 shows the U.S. Price and Wage Phillips Curve starting in 1945 to today, with data from The FederalReserve Bank of Saint louis.

8

Bretton−Woods System Flexible Exchange Rates

Phillips Curve Period

$ deval (Nixon)

Stagflation

$ deval (Plaza)

Missing InflationMissing Deflation

Missing Inflation−2%

0%

2%

4%

6%

8%

10%

12%

14%

45 50 55 60 65 70 75 80 85 90 95 00 05 10 15 20

Une

mpl

oym

ent R

ate,

Infla

tion

Inflation (Price)Inflation (Wage)Unemployment Rate

Figure 2: U.S. Price and Wage Phillips Curve (1945-)

Indeed, at that time, the U.S. Phillips curve was a very strong one. Table 10 shows that the wage Phillipscurve also was significant before 1971, but not after.

Table 10: U.S. Wage Phillips CurvesPeriod Exchange Rate Regime U.S. PC Slope t-stat Adj. R2

1948-1971 Bretton Woods -1.6*** -4.4 44%1971-2018 Flexible Exchange Rates 0.32 1.3 1.7%

CPI Components. Much more data is available for the United States, which allows to test the hypothesismore precisely. I now use the Bureau of Labor Statistics’ All Urban Consumers series (BLS-CU) in orderto investigate which components of the growth in the Consumer Price Index are negatively related tounemployment, and which are not. I test the original version of the phillips curve, that is, for each sector s, Itest whether price inflation in this sector is related to the unemployment rate in the time series:

πst = βsUt + εt

The Phillips curve coefficients βs from these regressions are reported in Table 11. For the sake of brievety,I report only the sectors for which the regression has a high explained variance is high (that is, wherethe adjusted R2 is higher than 20%). These results are presented as follows. The top panel shows thesectors where there is a negative relationship between price inflation and unemployment. They are rankedby decreasing order of adjusted R2. The bottom panel shows the sectors where the relationship is positive,contradicting the Phillips curve.

Apart from a few noteworthy exceptions, we can notice a pattern. Prices for which the local cost componentis important such as rents, local services (personal care, services), tend to have a strong Phillips curve. Thefact that “rent of shelter” has a strong negative relation to unemployment is very meaningful, as this itemrepresents more than 35% of the overall CPI index. On the other hand, manuyfacturing goods such as newcars, new motorcycles, which also represent a large fraction of total spending, tend to go opposite to thePhillips curve.

9

Table 11: U.S. Price Phillips Curves on CPI Components (Adjusted R2 higher than 20%)Item (U.S. City Average, All Urban consumers, SA) PC Slope t-stat Adj. R2

Other recreation services -0.51*** -15.3 49.1%Rent of shelter -0.42*** -17.4 47.3%

Information technology commodities -0.69*** -7.7 37%Club memberships -0.57*** -11.2 34%

Televisions -2.16*** -10.2 27.8%Moving, storage, freight expense -0.91*** -9.5 27%

Personal care -0.19*** -8.7 24.5%Admission to movies, theaters, and concerts -0.44*** -8.3 22.9%

Lodging away from home -0.81*** -8.1 21%Cigarettes -0.38*** -5.1 20.1%

Transportation commodities less motor fuel 0.65*** 11.1 55.6%Toys, games, hobbies and playground equipment 0.98*** 11.8 53%

Parking fees and tolls 0.96*** 9.3 50.5%Recreation commodities 0.28*** 8.4 41.3%

Water and sewerage maintenance 0.64*** 16.2 40.4%Water and sewer and trash collection services 0.41*** 12.0 37%

Education and communication services 0.45*** 7.4 35.2%New cars and trucks 0.44*** 11.0 33%New motorcycles 1.22*** 6.8 29%

Women’s underwear, nightwear, sportswear and accessories 0.79*** 9.4 26.7%College tuition and fees 0.8*** 12.9 25.7%

Medical care commodities 0.84*** 14.4 25.1%Lunchmeats 0.75*** 5.8 24.7%New trucks 0.67*** 10.7 21.7%

Inpatient hospital services 0.43*** 8.2 20.6%Other recreational goods 0.46*** 8.0 20.4%

Tuition, other school fees, and childcare 0.66*** 11.0 20%

1.4 United KingdomThe correlation between unemployment and inflation is usually attributed to A.W. Phillips, who is usuallybe thought to have been the first to document this relationship in Phillips (1958) for the United Kingdom.Phillips (1958) documents a negative relationship between wage inflation and unemployment, the wagePhillips Curve. Figure 3 shows Phillips’ original relationship between wage inflation and the unemploymentrate. On this graph, the correlation between unemployment and wage inflation can be seen even withoutrunning a regression: the R2 is high, the effects were significant.

Tables 12 and 13 extend A.W. Phillips’ work in the United Kingdom over different historical periods. Bothtables 12 and 13 show that the wage and price Phillips curve works in fixed exchange rate periods, but notunder flexible exchange rates. This period, unlike for the United States, starts in November 1967, when theU.K. devalues the pound by about 14% to $2.40, down from $2.80. Over the period starting in 1967 andending in 2016, the Phillips curve is no longer a feature of the data.

2 Event Studies: U.S. 1933 and 1971One could argue that countries under fixed versus flexible exchange rates are different in other respects thanjust the exchange rate regime. In that case, the previous differences in the strength of Phillips Curves underdifferent exchange rate systems can be due to unobserved country characteristic driving both the exchangerate regime and the existence of a Phillips curve.

10

Silver / Gold StandardSuspensions / Devaluations

Bretton Woodssuspend

returndeval

Original Phillips (1861−1957)Out−of−sample PhillipsHigh Nonlinearities

−28%−26%−24%−22%−20%−18%−16%−14%−12%−10%−8%−6%−4%−2%

0%2%4%6%8%

10%12%14%16%18%20%22%24%26%28%30%32%

1760 1770 1780 1790 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960

Une

mpl

oym

ent R

ate,

Infla

tion

Inflation (Wage)Unemployment Rate

Figure 3: U.K. Wage Phillips Curve (1760-1945): Unemployment and Wage Inflation.

Original Phillips(−1957)

Out−of−samplePhillips

Bretton Woods Flexible Ratesdeval.End Bretton Woods

Stagflation Missing Deflation−2%

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

22%

24%

26%

28%

50 60 70 80 90 00 10

Une

mpl

oym

ent R

ate,

Infla

tion

Inflation (Wage)Unemployment Rate

Figure 4: U.K. Wage Phillips Curve (1945-): Unemployment and Wage Inflation.

11

Table 12: U.K. Wage Phillips CurvesPeriod Exchange Rate Regime U.K. PC Slope t-stat Adj. R2

1780-1860 Silver / Gold Standard -0.33** -2.6 5.2%1861-1913 Gold Standard -0.79*** -7.1 48.4%1914-1925 Suspension -3.07*** -6.1 76.8%1926-1944 Gold Standard -0.71** -2.8 26%1945-1967 Bretton Woods -2.14*** -3.0 27.6%1967-2016 1967 Deval., Flexible Exchange Rate -0.47 -1.5 2.7%

Table 13: U.K. Price Phillips CurvesPeriod Exchange Rate Regime U.K. PC Slope t-stat Adj. R2

1780-1860 Silver / Gold Standard -0.73** -2.0 2.9%1861-1913 Gold Standard -0.63** -2.6 10.1%1914-1925 Suspension -3.02*** -8.6 86.9%1926-1944 Gold Standard -0.68*** -3.9 43%1945-1967 Bretton Woods -1.95* -1.9 11.7%1967-2016 1967 Deval., Flexible Exchange Rate -0.35 -1.2 0.9%

In this section, I use two examples of what Nakamura and Steinsson (2018) call “discontinuity basedidentification”. These two episodes are very salient in U.S. macroeconomic history. First, I look at whathappened to the Phillips Curve, when the United States devalued the dollar in 1933. Second, I consider thechange in the slope of the Phillips Curve around the time where the U.S. left of the Gold Standard in 1971.

2.1 Post 1933 U.S.: Missing DeflationIn their seminal paper on the U.S. Phillips curve, Samuelson and Solow (1960) were writing about theRoosevelt era “missing deflation”: “In the first place, the years from 1933 to 1941 appear to be sui generis:money wages rose or failed to fall in the face of massive unemployment.” After 1933, and until the UnitedStates entered World War II, deflation gave way to inflation despite high unemployment, including duringthe 1937-1938 recession. What is a “puzzle” according to the traditional Phillips curve view can be explainedvery well according to the version of the Phillips curve I previously laid out.

According to this interpretation, the failure of wages to fall during 1933-1941 was due to Roosevelt’s decisionin 1933 to devalue the dollar. The reason for the failure of the Phillips curve in the 1970s, and its failuresthereafter (missing inflation in the late 1990s and 2010s, missing deflation in 2007-2009), was not a change ininflation expectations, but coming from the fact that the U.S. were under flexible exchange rates startingfrom the end of the Bretton-Woods system in 1971. Therefore, the stagflation of the 1970s was no vindicationof “rational expectations”, as the story is usually told in textbooks, but more simply a manifestation thatthe structural relationship really is between real exchange rate growth and unemployment. The reason whyregional Phillips curves are always very strong, even when national Phillips curves fail, is that U.S. regionsare in a monetary union. To the best of my knowledge, the link between the stagflation period and the fall ofthe Bretton Woods System, the end of deflation in 1933 and Roosevelt’s devaluation of the dollar, or thedifference between regional and aggregate Phillips Curves based on their differences in exchange rate regimeshave not been noted before.

The Phillips Curve starts breaking in post-1933 Great Depression United States, which sees the coexistenceof price, wage inflation, as well as a substantial unemployment rate. This is usually considered a challenge forthe Phillips (1958) curve, as unemployment remained very high throughout the period. For instance, in aleading macroeconomics textbook, Blanchard (2016a) writes that “Starting in 1934, however, deflation gaveway to inflation, leading to a large decrease in the real interest rate, and the economy began to recover. Why,despite a very high unemployment rate, the U.S. economy was able to avoid further and further deflationremains a hotly debated issue in economics.”

12

100

150

200

250

300

350

400

20 23 26 29 3719 21 24 27 33 38 41

Pric

e Le

vel

Figure 5: Missing Deflation (1933-1941)

This text exemplifies that according to New-Keynesian economics, changes in the relationship betweeninflation and unemployment must necessarily arise from changes in inflation expectations. Therefore, thisis what changing the monetary system must have led to. On the contrary, in this paper I interpret thecoexistence of inflation and unemployment as a proof that the Phillips curve is really about real exchangerates. The devaluation of the dollar brought about the fall in real exchange rates, which was necessitated bythe crisis.

2.2 1970s Stagflation: the Triumph of Rational Expectations?In the 1970s, only a decade after the neoclassical synthesis was proposed, a period of high inflation andhigh unemployment shattered the Phillips Curve. The 1970s stagflation was considered a major challenge tomainstream Keynesian macroeconomic thinking at the time, which was based on the Phillips curve. Blanchardand Summers (2017) write: “A combination of intellectual developments and real world events led to adramatic reconceptualization of macroeconomics between the late 1960s and the early 1980s. Phelps (1967)and Friedman (1968) pointed out that, on theoretical grounds, one would not expect to see a stable tradeoffbetween inflation and unemployment as postulated by the simple Phillips curve, as agents’ expectationswould change if policymakers tried to exploit the Phillips curve too much. At the same time by the late1970s, and in apparent contrast to the Keynesian view, stagflation emerged as a major problem throughoutadvanced economies as inflation and unemployment both increased in unison.” This disappointment led to afierce attack of neoclassical economics against Keynesian economics (Lucas (1973), Lucas and Sargent (1979)),which continues until today.

Accelerationist Phillips Curve models came to replace the original Phillips Curve, following Phelps (1967) andFriedman (1968): in his 1968 presidential address, Milton Friedman argues that “There is always a temporarytrade-off between inflation and unemployment; there is no permanent trade-off.” Phelps (1967) and Friedman(1968) argued that the trade-off could not be persistently exploited, as people’s expectations would change.The early correlation between inflation and unemployment was replaced with the expectations-augmentedPhillips curve, a correlation between inflation growth and unemployment. Yet even this version of the PhillipsCurve quickly had problems too, so much that Larry Summers was asking in 1991 whether “Keynesianeconomics [should] dispense with the Phillips Curve” (Summers (1991)). There was no rise in inflation during

13

the late 1990s despite a strong economy, and Atkeson and Ohanian (2001) showed that Phillips Curveswere not on average useful to forecast inflation. The 2007-2009 crisis and subsequent recovery did not occuraccording to the Phillips Curve either: there was no desinflation in 2007-2009 despite an unprecedentedrecession since the Great Depression, and an unemployment rate rising from 4.5% of the labor force in April2007 to 10% in October 2009. Referring to this period, Hall (2011) talked about the “near-exogeneity of therate of inflation”. Symmetrically, there has been no sign of rising inflation during the recovery, at least until2019, despite unprecedented tightness in the labor market. Janet Yellen (2017) has called it “the biggestsurprise in the U.S. economy.” Looking back on the performance of macroeconomics since the financial crisis,Paul Krugman (2018) has argued recently that there exists “a big failure in our understanding of pricedynamics.” At the same time, regional Phillips Curves are always very strong: there is much evidence that ahigh correlation exists between local unemployment and local inflation across U.S. regions, even in periodswhen the Phillips Curve fails at the aggregate level (Beraja, Hurst, and Ospina (2016), Hooper, Mishkin, andSufi (2019)).

At the same time, the expectations-augmented Phillips curve seemed to still be a reasonably good fit to thedata: instead of a correlation between unemployment and inflation, there was now a correlation betweenunemployment and the growth in inflation. This episode is usually interpreted as one of a rare case of “theoryahead of facts” on the part of Phelps (1967) and Friedman (1968), who had anticipated based on rationalexpectation theorizing that the Phillips curve trade-off could not be exploited for too long (see Blanchard(2016a) for a textbook treatment). The usual narrative is that Phelps (1967) and Friedman (1968) wereproven right by the 1970s stagflation, which led the Phillips curve to be replaced by the accelerationistPhillips Curve: a negative relationship between the change in inflation and unemployment. This alleged“success” also led to the triumph of rational expectations economics.

This paper proposes another interpretation of the breaking of the Phillips curve in the 1970s: Nixon leftthe Gold standard, which ended the Bretton Woods system of fixed exchange rates. Therefore, the U.S.economy moved from having a fixed exchange rate with a strong price Phillips curve (which was also areal exchange rate Phillips curve), to a flexible exchange rate where the correlation would now be betweennominal exchange rate and unemployment, instead of between price inflation and unemployment. In thatinterpretation, the breaking of the Phillips curve should not have been interpreted as a triumph of rationalexpectations economics, or as a problem for Keynesian economics, but simply as a consequence of a differentexchange rate regime.

What about the accelerationist Phillips curve? Although the accelerationist version of the Phillips Curve(between the changes in inflation and unemployment) replaced the Phillips Curve, it in fact always was noisy,and received numerous justified critiques (Atkeson and Ohanian (2001)). Recently, the accelerationist versionof the Phillips curve has not done very well either, so much that Blanchard (2016b) has proposed to go backto the old version of the Phillips curve.

3 Large Shocks: U.S. Missing Deflation and Missing InflationAs stated by Nakamura and Steinsson (2018), “Much empirical work takes the approach of seeking to controlfor confounding factors as well as possible. A different approach is to focus on large policy actions for whichwe can plausibly argue that confounding factors are drowned out.” In this section, I argue that both the2007-2009 financial crisis, as well as the 2016-2019 stimulus when the unemployment rate was already low aretwo such episodes. Both events have been characterized by absentee Phillips curve, but very large responses ofrents and house prices. I argue that both these events are also supportive that a structural relationship existsbetween real exchange rates and unemployment, but not between overall cpi inflation and unemployment.

3.1 2007-2009 Missing DeflationIf the Phillips curve view of the world is correct, then the 2007-2009 financial crisis, which was the mostsevere since the Great Depression, should have led to substantially more deflation. This missing deflation is avery puzzle problem for the Phillips curve. Among many examples, Hall (2011) writes: “A line of thoughtrather deeply embedded in macroeconomics holds that product prices fall in slack markets. (. . . ) Recent

14

experience requires a fundamental reconsideration of the view that producers find it desirable to expandoutput by cutting prices. Their behavior across all industries suggests, to the contrary, that price-cutting isnot the answer to any problem they perceive in a time of extreme slack.” Apart from a few exceptions, mostcommentators consider the Phillips curve to have been a failure.4 Because of missing desinflation duringthe Great Recession, Hall (2011) has talked about the “near-exogeneity of the rate of inflation”. Even PaulKrugman (2018) has argued that “So my claim that basic macroeconomics worked very well after the crisisneeds to be qualified by what looks like a big failure in our understanding of price dynamics — but thisfailure didn’t do too much damage in giving rise to bad advice, and hasn’t led to big new ideas becausenobody seems to have good ideas to offer.”

The behavior of inflation during the financial crisis of 2007-2009, as well as during the ongoing recovery, alsovery hard to understand through the lens of the Phillips Curve.

3.2 2016-2019 Missing InflationSymmetrically, inflation has not picked up since the recovery started, which Janet Yellen has called “thebiggest surprise in the U.S. economy” (Yellen (2017)).

4 Identified MomentsOf course, one cannot identify structural relationships with correlations only, as in general supply and demandshocks shift both supply and demand curves, moving along demand and supply curves respectively. Therefore,looking at simple correlations such as the relationship between inflation and unemployment, and trying touncover an aggregate supply curve, the implicit assumption is that most shocks are aggregate demand shocks,so that the resulting correlation allows to trace an aggregate supply curve. An alternative to event studies isto stack many episodes where aggregate demand is known to have increased, and compute the (conditional)covariance between unemployment and inflation during these episodes. Again, the response to both fiscal andmonetary policy shocks strongly suggests that the structural aggregate supply curve is a relation betweenreal exchange rates and unemployment, instead of inflation and unemployment. I show that in the U.S.,unemployment rises strongly following a contractionary fiscal or monetary shock, while the Consumer PriceIndex responds ambiguously. However, the relative price of housing falls substantially following both suchshocks.

4.1 U.S. Monetary Shocks (Romer and Romer (2004))I start with monetary shocks identified in Romer and Romer (2004). In the New-Keynesian literature,monetary shocks are indeed usually identified as pure aggregate demand shocks. I estimate the simplestpossible specification:

∆Yt = a+M∑i=0

bi∆Tt−i + et

Here, I set the number of lags to M = 20. The results from the cumulated partial responses bi are shown onFigures 6, 7, and 8, together with the 68% and 90% confidence intervals obtained by bootstrap. We note thatfollowing a monetary shock, unemployment rises as shown on Figure 6. At the same time, the price levelshows a “price puzzle”: the price level first appears to rise, and only after two years to start to decline, asshown on Figure 7. Finally, Figure 8 shows that house prices unambiguously decline starting in quarter 1.Once again, the Phillips curve pattern is much more apparent on house prices than it is on general priceinflation.

4.2 U.S. Fiscal Shocks (Romer and Romer (2010))I next use the narrative shocks identified in the United States by Romer and Romer (2010). Using theirmethodology, I compute the impulse response functions to a 1% of GDP increase in taxes. Figure 9 shows

4Blanchard (2018) however is more nuanced on how one should interpret the missing deflation.

15

−12%

−10%

−8%

−6%

−4%

−2%

0%

2%

4%

6%

8%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 6: Response of Unemployment to a 1% Increase in the Federal Funds Rate

−4%

−3%

−2%

−1%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 7: Response of CPI Inflation to a 1% Increase in the Federal Funds Rate

16

−9%

−8%

−7%

−6%

−5%

−4%

−3%

−2%

−1%

0%

1%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 8: Response of House Prices to a 1% Increase in the Federal Funds Rate

that conditional on such shocks, unemployment increases. Figure 10 shows that inflation does not movesignificantly, although a Phillips curve type analysis would suggest it should. Finally, Figure 11 shows thatsimilarly to monetary policy shocks, house prices respond much more than the overall Consumer Price Index.Again, this evidence is supportive with the main thesis of the paper: even conditionally on identified aggregatedemand shocks, inflation does not respond in the way that the Phillips curve would say it should, while onthe contrary house prices, the real exchange rate, or the relative price of non-tradable goods in terms oftradables, do.

Compared to monetary shocks, fiscal shocks are also sometimes interpreted as arising from supply responses,and not just from a change in disposable income which boosts consumption. However, it is also well-knownthat it is hard to make sense of the large effects of tax shocks on output, since the implied elasticities would bean order of magnitude higher than the much better identified microeconomic elasticities. Therefore, aggregatedemand effects must be large.

5 Regional Phillips CurvesThe correlation between unemployment and local price indexes is already well-known in the literature. Morerecently, Hooper, Mishkin, and Sufi (2019) have studied these correlations more systematically. However, tothe best of my knowledge, all these studies use cross-sectional data in order to say something about Phillipscurves at the aggregate level. None of these studies have noted that a correlation between regions in amonetary union was equally compatible where the structural relationship is between unemployment andexchange rates. I study the different geographies in turn.

5.1 U.S. StatesThe fact that nominal wage growth and unemployment are correlated at the state level is well-known at leastsince Blanchard and Katz (1992). They perform a VAR analysis, and compute the impulse response functionto an “employment shock”). In response to an employment shock, they get a fall in nominal wages, as well as

17

−2%

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 9: Response of Unemployment to a 1% of GDP Tax Increase

−2%

−1%

0%

1%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 10: Response of CPI Inflation to a 1% of GDP Tax Increase

18

−14%

−12%

−10%

−8%

−6%

−4%

−2%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 11: Response of House Prices to a 1% of GDP Tax Increase

a fall in employment, which corresponds to a Phillips curve.

More recently, Zidar (2019) has investigated the state level impact of Romer and Romer (2010)’s fiscalshocks, depending on whether tax increases are concentrated on the Bottom 90% of earners (comprisingapproximately 50% of the income), or the Top 10%. Again, these impulses are potentially more informative,as they are conditioned on an identified fiscal shock; although just as for Romer and Romer (2010)’s shocks,however, there can be a debate on whether these shocks are to aggregate demand or aggregate supply.5

The results are nonetheless instructive, as he finds that prices tend to fall, as shown on Figure 13B, whilecomposition-adjusted real wages stay constant, as shown on Figure 14C. Again, this implies a nominal wagePhillips curve, as well as a price Phillips curve. Once again, fiscal austerity drives inflation down, mostlythrough the price of housing.

Table 14: House Prices and Unemployment (Cross-sectional Regressions, State Level)Period PC Slope t-stat Adj. R2

1990-1995 -7.1*** -3.9 22%1999-2005 -6.06** -2.4 9%2006-2009 -6.98*** -5.8 40%

5.2 U.S. MSAsSince house prices probably vary most significantly at the city level, I now run similar cross-sectionalregressions for U.S. Metropolitan Statistical Areas (MSAs). Again, regardless of the housing cycle one looksat, there exists a strong correlation between unemployment and house price growth, as shown on Table 15.

5And indeed, Zidar (2019) states: “I find that real wages increase after tax changes for lower-income groups. While theestimates are imprecise, they suggest that labor supply responses are an important mechanism for the results.” It should benoted however that all the impulse responses studied in this paper are very similar, while they typically correspond to aggregatedemand shocks.

19

Olivier Jean Blanchard and Lawrence F. Katz 41

Figure 12. Response of Employment and Adjusted Wages to an Employment Shock Effect of shock (percent)

0

-0.25

-0.50-

-0.75-

SO f X >t Emp~~~~~~~~~~loyment

-1.00

-1.25-

-1.50-

-1.75 -

-2.00 _ I I I I I I

0 2 4 6 8 10 12 Year

Source: Authors' calculations using data described in the appendix. The shock is a - I percent shock to employment. Bands of one standard error are shown around each line.

years, thus implying an elasticity of relative wages to relative unemployment of approximately unity.44

While the manufacturing wage is the only wage series available con- sistently for the last 40 years, there are reasons to think that it may be a mediocre proxy for movements in overall wages in a state. Thus we have constructed another wage series for the period 1979-89, by using CPS data to construct a wage for each state and each year controlling for a number of industry and workers characteristics. (The details of construction appear in the appendix.) Figure 12 gives the dynamic responses of employment when we use that wage instead of the manufacturing wage, doing panel data estimation allowing for state effects. In this case, the time dimension is short so that even when we use only two lags on each variable, we have only six degrees of freedom for each state. Thus while the pooling of all states gives us substantial information at short horizons, our estimated responses are unlikely to be accu-

44. These findings are consistent with the general wisdom from past research. For a survey of earlier research and extensions, see Bartik (1991).

Figure 12: Reponse of Employment and Adjusted Wages to an Employment Shock (Blanchardand Katz (1992))

20

-.15

-.1-.0

50

.05

.1P

erce

nt

-3 -2 -1 0 1 2 3 4Years Since Tax Change

A. Nominal GDP

-.15

-.1-.0

50

.05

.1P

erce

nt


B. ACCRA Price Index

-.2-.1

0.1

.2P

erce

nt


C. Real GDP (ACCRA)

-.15

-.1-.0

50

.05

.1P

erce

nt


D. Real GDP (Moretti HPI)

Bottom 90 Top 10Figure 13: Nominal GDP, Real GDP, Price Indexes Responses to a 1% of GDP Shock (Red:Top 10%; Blue: Bottom 90%). Source: Zidar (2019).

21

-.1-.0

50

.05

Per

cent

age

poin

ts


A. Labor Force Participation Rate

-.04

-.02

0.0

2.0

4P

erce

nt


B. Hours-.2

-.10

.1.2

Per

cent


C. Composition-Constant Real Wages (ACCRA)

-.15

-.1-.0

50

.05

.1P

erce

nt


D. Consumption

Bottom 90 Top 10Figure 14: Participation, Hours, Real Wages, Consumption Responses to a 1% of GDP Shock(Red: Top 10%; Blue: Bottom 90%). Source: Zidar (2019).

Alabama

Alaska

Arizona

Arkansas

California

ColoradoConnecticut

DelawareDistrict of Columbia

Florida

GeorgiaHawaii

Idaho

Illinois

Indiana

IowaKansas

Kentucky

Louisiana

Maine

MarylandMassachusetts

Michigan

Minnesota

Mississippi

Missouri

Montana

Nebraska

Nevada

New HampshireNew Jersey

New Mexico

New York

North Carolina

North Dakota

Ohio

Oklahoma

Oregon

Pennsylvania

Rhode Island

South CarolinaSouth Dakota

Tennessee

TexasUtah

Vermont

Virginia

Washington

West Virginia

Wisconsin

Wyoming

−60%

−50%

−40%

−30%

−20%

−10%

0%

10%

0% 1% 2% 3% 4% 5%∆ Unemployment Rate 06−09

∆ H

ouse

Pric

es 0

6−09

Figure 15: U.S. States Phillips Curve (06-09)

22

Table 15: House Prices and Unemployment (Cross-sectional Regressions, MSA Level)Period PC Slope t-stat Adj. R2

1990-1995 -2.34*** -7.2 22%1999-2005 -4.14*** -8.3 9%2006-2009 -5.48*** -13.7 40%

This correlation is plotted on Figure 16 for the 2006-2009 recession. For example, Merced, California had ahigher increase in the unemployment rate and lower house prices price inflation than Midland, Texas.

Cape Coral−Fort Myers FL

Gulfport−Biloxi−Pascagoula MS

Merced CA

Midland TX

Modesto CA

Odessa TX

Salinas CA

Stockton−Lodi CA

Vallejo−Fairfield CA

−100%

−90%

−80%

−70%

−60%

−50%

−40%

−30%

−20%

−10%

0%

10%

20%

30%

40%

50%

60%

−8% −6% −4% −2% 0% 2% 4% 6% 8% 10% 12% 14%∆ Unemployment Rate 06−09

∆ H

ouse

Pric

es 0

6−09

Figure 16: U.S. MSAs Phillips Curve (06-09)

5.3 Euro Area CountriesAnother example in support of this hypothesis is that of Euro area countries during the early 2000s, whereperiphery countries had higher consumption, trade deficits, massive capital inflows, higher price inflation,together with a booming economy. (Martin and Philippon (2017)) This pattern was reversed in the bust, asshown on Figure 17. There was deflation in Greece, associated with high uneployment, following the collapsein aggregate demand prompted by fiscal austerity. Therefore, there was a strong correlation both in theboom and in the bust, between inflation and economic activity. At the same time, according to the externalevidence provided above, this should on be interpreted as a structural relationship between competitivenessand economic activity, instead of a traditional Phillips curve.

One often accepted narrative is that fiscal policy was overall too lax in periphery countries such as Greece andPortugal, and that Spain and Ireland did not do enough to contain private leverage. Sections ?? and ?? discussthese important issues. While it is true that expansionary fiscal policy undertaken without coordination runsinto potential external balance problems, it is also true that there exists an excess of aggregate savings at theworld level. As a consequence, the problem was as much that periphery countries were doing too lax policiesthat core countries such as Germany were not doing the same.

23

Austria

Belgium

GermanySpain

Estonia

Finland

France

Greece

Ireland

Italy

Lithuania

Luxembourg

Latvia

Netherlands

Portugal

Slovenia

−2%

−1%

0%

1%

2%

3%

4%

5%

6%

7%

8%

4% 6% 8% 10% 12% 14% 16% 18% 20%Unemployment Rate (2011)

Infla

tion

(201

1−20

15)

Figure 17: Euro Area Correlation between Inflation and Unemployment (2011-2015)

6 Implications of the Real Exchange Rate Phillips CurveIn the previous sections, I have provided evidence in favor of the view that the Phillips curve is not acorrelation between price or wage inflation and unemployment, but that it is instead a robust correlationbetween real exchange rate growth and unemployment, or equivalently a correlation between the relativeprice of non-traded goods (mostly housing and rents) over traded goods and unemployment. What are theconsequences of this finding for positive and normative analysis? I first show that this finding is importantfor identifying the source of business cycle fluctuations, and telling apart between aggregate demand andaggregate supply shocks. This is important for central banks and other actors interested in questions ofmacroeconomic stabilization. Second, I argue that the costs of “too high” aggregate demand are perhapsmore found in an “overvalued” real exchange rate, than in the (small) welfare costs of inflation. However,I emphasize that under secular stagnation, the adjustment should fall on countries with trade surplusesrather than on those which have trade deficits. Third, I show that the trade-off between real exchange ratesand unemployment is actually already well-known to policymakers and monitored by them. However, manyeconomists are often skeptical of these so-called competitiveness concerns, so that the real exchange ratePhillips curve might help bridge the gap between academics and practioners. Fourth and finally, I show thatthe real exchange rate Phillips curve leads to a different assessment of Keynesian economics. I only sketchthe surface here, in depth analysis of these issues is left for future research.

6.1 Identifying the source of business cycle fluctuationsAccording to the real exchange rate Phillips curve, real exchange rates are a better measure of aggregatedemand conditions than inflation. This has important consequences for identifying the source of businesscycle fluctuations. For example, high aggregate demand can come together with overall low CPI inflation, ifthe nominal exchange rate appreciates at the same time. A sign of this phenomenon is that house pricesare then relatively high, even though inflation is lower because imported goods are cheaper. Conversely,deficient aggregate demand will not necessarily lead to deflation or falling inflation, if the nominal exchangerate depreciates by enough.

24

High aggregate demand was accompanied by low inflation at the end of the 1990s, when Alan Greenspan’sterm as a Federal Reserve chairman was accompanied with “missing inflation” despite a booming economy.The increase in housing prices started at that time, and continued throughout the housing boom of the 2000s.Alan Greenspan was actually even blamed for having made interest rates low for too long, which fuelled thehousing bubble. Similarly, inflation has not risen in the United States recently, despite an unprecedentedfiscal stimulus in an already low unemployment economy. More generally, with a real exchange rate Phillipscurve, an appreciating dollar implies that aggregate demand shocks need not be inflationary, and that theFederal Reserve should monitor at the relative price of non-traded goods versus traded goods, if the objectiveis aggregate demand stabilization, and avoiding “overheating.” I however discuss in the next section that“overheating” is only relative, and that under secular stagnation, the burden of the adjustment should fall onsurplus countries, not on a deficit countries, because of deficient global aggregate demand.

Symmetrically, low aggregate demand was accompanied by relatively high inflation in the famous 1970sepisode of the “stagflation.” This was no failure of Keynesian economics, or the proof of a supply shock, asthe story is usually told in undergraduate textbooks, but rather coming from the fact that the dollar haddepreciated substantially following Nixon’s decision to end the Bretton-Woods system of fixed exchange rates.Similarly, one could infer from the “missing deflation” between 2007 and 2009 that the U.S. simultaneouslyexperienced an aggregate demand and an aggregate supply shock, which has been argued by some scholars(Beraja, Hurst, and Ospina (2016)). According to the real exchange rate Phillips curve, this can perfectly beexplained by the fact that all countries experienced a similarly sized negative aggregate demand shock, sothat the real exchange rate did not need to adjust by much. And even if the U.S. has been disproportionatelyhit by the financial crisis, the dollar could have depreciated, which would have replaced deflation in the realexchange rate adjustment. Again, this failure of prices to fall are usually considered a puzzle for Keynesianeconomics: Paul Krugman (2018) has recently argued that even though macroeconomic analysis fared wellduring the crisis, there exists “a big failure in our understanding of price dynamics.”

More generally, the real exchange rate Phillips curve should allow to better account for the relative importanceof aggregate supply and demand shocks to macroeconomic fluctuations. Given the very strong real exchangerate Phillips curve as given by the significance of the previous regressions, or their explanatory power (highR2), accounting for business cycles through the lens of the real exchange rate Phillips curve will likely give avery important contribution to fluctuations in aggregate demand. However, a full identification the source ofbusiness cycle fluctuations based on the real exchange rate Phillips curve is left to future research.

6.2 The costs of overheatingIn the new-Keynesian model, the costs of overstimulating the economy are found in the welfare cost ofinflation, coming from the dispersion in prices. However, Nakamura et al. (2018) have shown based onempirical evidence on price dispersion during the Great Inflation of the late 1970s and early 1980s that “thestandard New Keynesian analysis of the welfare costs of inflation is wrong”, since the welfare costs of inflationare way too small.

If the Phillips curve does not represent a trade-off between inflation and unemployment, but rather one betweenreal exchange rates and unemployment, then the costs of “overstimulating” are potentially very different.The associated costs of high aggregate demand are represented by an oversized real estate, construction, andnon-traded sector, which can be detrimental to growth (Rodrik (2008)). In the next section, I discuss thatthis potential cost is in fact very much discussed by policymakers already, and potentially very salient.

The notion of an “overheating” economy in the context of a real exchange rate Phillips curve should be furtherdiscussed, however. With demand-side secular stagnation (Geerolf (2013b), Geerolf (2019)), the problem liesmore in low aggregate demand in surplus countries than in too high aggregate demand in deficit countries:there is a global excess of saving over investment, and surplus countries simply “export their way out” of thisproblem, by finding stores of value abroad. In fact, these issues of surplus countries’ adjustment was centralto John Maynard Keynes’ thinking around the creation of the International Monetary Fund: this was theso-called Keynes plan at Bretton Woods.6

6See Steil (2013) for historical background around the Battle of Bretton Woods.

25

6.3 A well-known policy trade-off: aggregate demand and competitivenessThe trade-off between unemployment and real exchange rate growth, corresponding to the real exchangerate Phillips curve, is in fact well-known both in the academic as well as the policy world, where issuesof external imbalances are often discussed and monitored, albeit usually independantly from inflationarypressures. These worries are often behind the reluctance to stimulate aggregate demand, even when inflationis low. The contribution of this paper is to show that the external balance constraint, and worries aboutinflation, in fact represent the two sides of the same trade-off: the real exchange rate Phillips curve. Thisworry is known both in the case of emerging markets, as well as in advanced economies.

The case of emerging markets. Although it is not usually associated to the Phillips Curve, a vast bodyof research has studied the potential effects of aggregate demand stimulating policies on the real exchangerate and competitiveness. In fact, it is well known in business cycle research of emerging economies, thatthey follow strong patterns of boom-bust episodes. In the boom, there is a rise in the real exchange rate, amove of the economy towards the non-traded sector, a deterioration in competitiveness, together with a risein consumption (Kalantzis (2015)). For example, Rebelo and Vegh (1995) describe how disinflation camefrom the stabilization programs in the Southern Cone of Latin America (Argentina, Chile, and Uruguay)in the late 1970s, and the real effects of these experiences: “countries that use the exchange rate as thenominal anchor in inflation stabilization programs experience a boom in economic activity (consumption,investment, and GDP expand), a large real exchange-rate appreciation, a rise in the real wage rate, anda deterioration in the external accounts. Later in the programs, these effects are often reversed, with theeconomy contracting sharply and the real exchange rate depreciating.” Again, this is just a reflection ofthe real exchange rate Phillips curve. In fact, this is often a very important argument against Keynesian,stimulative policies, which is given in the policy world. A textbook treatment of these effects is given in Vegh(2013), and in Schmitt-Grohé and Uribe (2016). For example, Figure 60 is taken from Vegh (2013), and itshows that boom busts episodes are characterized by a high relative price of non tradables during the boom,associated with a large current account deficit. I next show that this trade-off is in fact not only present inemerging markets, but that it is a well-known issue in advanced economies as well.

The trade-off in advanced economies. In fact, the suggested trade-off between unemployment and realexchange rates is not just relevant for emerging markets, it has in fact long been discussed by policymakersaround the world. The starkest example is provided by a policy publication called Economic Report of thePresident. For example, the 1963 Economic Report of the President report has a chapter called “Prices,Wages, and the balance of payments” (Kennedy (1963)). In this chapter, one can read the following: “Stabilityof prices is particularly important for the balance of payments. It should be emphasized, however, that whatis significant for America’s competitive position in international grade is not the absolute change in the levelof U.S. prices, but rather the change relative to prices abroad.” Current discussions around U.S. trade deficits,prompted by very accommodative fiscal policies, can also been understood in terms of the trade-off betweenreal exchange rates and unemployment. The Trump tax cuts have led to a nominal appreciation of the dollar,instead of price inflation in the United States. This might explain why inflation has not risen so far despiteunprecedented tightness in the labor market, which Janet Yallen has called “the biggest surprise in the U.S.economy.” (Yellen (2017))

Another example is provided by the United Kingdom, where aggregate demand stimulating policies areusually called “income policies”. In reviewing Forder (2014)’s critical take on the Phillips curve, Goodhart(2018) argues that “under Bretton Woods, the relevant trade-off in the U.K. was between growth and theBalance of Payments.” This appears to be a very general pattern, at least in fixed exchange rate regimes.Again, this trade-off is not unknown from policymakers, and even academics. For example, J.M. advocatedin favor of a general tariff on imports “without discriminating protective taxes” in Proposals for a RevenueTariff, published in March 1931, precisely to avoid the negative effects of stimulus policies on the balance oftrade.7

7“I do not believe that a wise and prudent Budget can be framed today without recourse to a revenue tariff. (. . . ) In sofar as it leads to the substitution of home-produced goods for goods previously imported, it will increase employment in thiscountry. At the same time, by relieving the pressure on the balance of trade it will provide a much-needed margin to pay for theadditional imports which a policy of expansion will require and to finance loans by London to necessitous debtor countries. Inthese ways, the buying power which we take away from the rest of the world by restricting certain imports we shall restore to itwith the other hand. Some fanatical Free Traders might allege that the adverse effect of import duties on our exports would

26

Another example is that of countries in the euro area, or even before the euro of countries inside the ExchangeRate Mechanism, as they were preparing for a common currency. For example, after experiencing withKeynesian policies at the beginning of François Mitterand’s first presidential term from 1981 to 1983, Franceembarked on austerity measures in 1983. Jacques Delors, then Minister of Finance, justified the fiscalconsolidation measures by referring to the growing French trade deficit (March 25 1983): “We cannot continueto consume more than we produce, to buy more than we sell abroad. For three, four years, France is in thissituation. This must change, and fast. . . We designed these measures as much as possible by reducing publicdeficits, and the least possible by directly reducing household incomes. . . This effort is only temporary. Itmust be massive enough to allow the rapid decline in imports in an open economy without protectionism.”Throughout the 1980s and 1990s, France was then said to be experiencing with “competitive desinflation”, asthe “franc” was trying to follow the “mark”: inflation and competitiveness were then seen as going hand inhand. Inflation was a problem not per se (because of price dispersion, for example) but rather because itmeant loss of competitiveness. In discussing Blanchard and Giavazzi (2002)’s optimistic assessment of tradedeficits in Europe, Pierre-Olivier Gourinchas was asking: “Should we worry?”, comparing the situation ofPortugal and Greece to that of Latin American countries which had adopted exchange rate stabilizationprograms.8 Once again, the relevant trade-off seemed to be between fiscal profiligacy, and trade deficits,rather than inflation.

Finally, the Japanese episode can be interpreted through the same theoretical framework. Again, the level ofthe real exchange rate played an important role. In 1985, the Plaza Accord forced Japan to move from anexport-driven model of growth, to a demand-driven model. This is reflected on Figure 59. An increase intrade deficits, together with a large increase in house prices and the price of land, ensued. Unemploymentrates reached a low point. In the early 1990s, this consumption-led boom came to a halt, unemploymentincreased, and the real exchange rate depreciated very sharply.

ConclusionSleeman (2011) in a retrospective on the Phillips curve argues that “A.W. Phillips was not satisfied withthe paper and had not intended to publish it in 1958.” He further notes: “I believe that Phillips was wellaware that his case for the stability of his famous curve was not compelling. His comments on his 1958 paperwere often dismissive: he referred to the paper as “a very crude attempt” (Leeson, 2000, p. 218), a “quickand dirty job” (Schwier, 2000, p. 24), something “just done in a weekend” (letter from Gregory quoted byLeeson, 1994a, p. 613), and as a “rushed job” (Blyth, 1978, p. xvi). The then-editor of Economica, BasilYamey (Leeson, 2000, p. 337), wrote that he would show Phillips the Phillips-curve papers submitted toEconomica, but that Phillips “always declined my invitation to write comments on the more substantialpieces or indeed to write a follow-up article to include his further reflflections. He seems to have lost interestin the subject soon after the paper was published. His fertile mind had moved on to other matters.” Holt, afellow engineer turned economist, who spent a sabbatical at LSE observes (2000, pp. 309–310): “I think thathe was a little embarrassed by the attention that the paper received../.. perhaps because both the empiricaleconometric work and the theory were conspicuously sloppy.”“. Even so,”the concept became familiar tostudents and teachers of economics a year later when Samuelson (1961, p. 383) incorporated the Phillipscurve in its trade-off form into the fifth edition of Economics, the textbook that dominated the teaching ofintroductory economics on both sides of the Atlantic in the early 1960s."

neutralise all this; but it would not be true.” Of course, because he was in favor of the international division of labor based oncomparative advantage, he advocated a tariff on the broadest possible range of goods: “The tariff which I have in mind wouldinclude no discriminating protective taxes, but would cover as wide a field as possible at a flat rate or perhaps two flat rates,each applicable to wide categories of goods”. For this same reason, his views were also quite sympathetic to mercantilism inChapter 23 of the General Theory (Keynes (1936)).

8“The experience of these two countries—up to this point—is very reminiscent of that of many Latin American countries thathave adopted exchange rate–based stabilization programs. Stabilization of the exchange rate, renewed access to internationalcapital markets, and some euphoria at the prospect of steady future growth combined to generate a strong consumptionboom—that is, a decline in saving—which may or may not have been accompanied by an investment boom. Growth was initiallysolid and everything looked benign. Over time, however, clouds gathered on the horizon: the currency appreciated in real terms,competitiveness plummeted, and foreign investors became worried as growth performance failed to meet expectations. Theendgame is well known: with a fixed exchange rate, restoring competitiveness required an adjustment in relative prices. Oftenthis was too little and too late. Eventually capital pulled out, forcing a devaluation.” (Blanchard and Giavazzi (2002))

27

Dissatisfaction on the Phillips curve has been expressed by prominent Keynesians before. In 1991, LarrySummers did not mince his words: “Frequent ad hoc adjustments to account for embarrassing realities were ahallmark of Ptolemaic astronomy. It is sad but true that the half-life of various Keynesian views about theaggregate supply curve has been little more than a decade. In The General Theory (1947) Keynes proposedthat the aggregate supply curve drawn in unemployment-price space was L-shaped. This view was falsifiedby the coincidence of inflation and less than full employment in the late 1940s and 1950s. By the early1960s, a derivative was slipped and Keynes’s view had given way to the Phillips curve vision of a stabledownward-sloping relationship between unemployment and the rate of inflation. This view remained popularfor not much more than a decade. The stagflation of the 1970s led to the slipping of another derivative andthe widespread acceptance of the view that there existed a natural rate of unemployment, which was the onlyrate at which inflation could remain stable. On this ‘expectations augmented’ Phillips curve view, there isa trade-off not between current inflation and current unemployment but between permanent inflation andcurrent unemployment.” According to him, “even its friends must acknowledge that the textbook Keynesianview of aggregate supply possesses many of the attributes that Thomas Kuhn has ascribed to dying scientificparadigms.” (Summers (1991)) In 2017, he reiterated: “Recent events are as severe a challenge to currentorthodoxy as the Depression was to the orthodoxy of John Maynard Keynes’ times or the inflation of the1970s was to the orthodoxy of its time.” (Summers (2017))

In this paper, I have argued that the Phillips correlation between inflation and unemployment perhaps neverwas about inflation, but that it was instead about real exchange rates. And in fact, Nakamura et al. (2018)have shown based on empirical evidence on price dispersion during the Great Inflation of the late 1970s andearly 1980s that “the standard New Keynesian analysis of the welfare costs of inflation is wrong”, since thewelfare costs of inflation are way too small. The results in this paper suggest that the impact of stimulativepolicies on competitiveness and trade deficits is potentially large. I have shown that although the failureof the Phillips Curve contradicts the New-Keynesian theory of Aggregate Supply, it is consistent with ademand-based secular stagnation view based on excess saving and limited investment needs. The empiricalevidence for such a “savings glut” has been illustrated in Geerolf (2013a) and Geerolf (2013b). A theoreticalmodel with demand-driven secular stagnation has been developed in Geerolf (2019). In fact, there was verylittle mention of sticky prices or wages in Keynes’ work, as argued very early on by Axel Leijonhufvud (1967)and others.

Is this view is correct, then the Phillips curve still might represent some meaningful trade-off, perhaps notbetween inflation and unemployment, but between competitiveness (real exchange rates) and unemploymentinstead. Indeed, one important concern with Keynesian policies that they have a negative effect on the balanceof trade, a concern which is much discussed in policymaking. In fact, concerted action of aggregate-demandstimulating policies is often undertaken on these grounds. Similarly, one can view the current trade tensionsbetween the United States and surplus countries (Germany and Japan) also in light of this issue. The mostpreferable course of action to deal with these problems is international coordination of fiscal policies. However,when such coordination is not possible, there is a trade-off for a given country taken in isolation. How tobest manage that trade-off between competitiveness and unemployment, through exchange rate policy, tradepolicy, or management of capital flows (which the IMF has recently argued in favor of), is an importantquestion for future research.

28

ReferencesAtkeson, Andy, and Lee E. Ohanian. 2001. “Are Phillips Curves Useful for Forecasting Inflation ?” FederalReserve Bank Minneapolis Quarterly Review 25 (1): 2–11.

Backus, David K., and Gregor W. Smith. 1993. “Consumption and Real Exchange Rates in DynamicEconomies with Non-Traded Goods.” Journal of International Economics 35 (3): 297–316. https://doi.org/10.1016/0022-1996(93)90021-O.

Beraja, Martin, Erik Hurst, and Juan Ospina. 2016. “The Aggregate Implications of Regional BusinessCycles.” Working Paper 21956. National Bureau of Economic Research. https://doi.org/10.3386/w21956.

Blanchard, Olivier. 2016a. Macroeconomics. Pearson Education.

———. 2016b. “The Phillips Curve: Back to the ’60s?” American Economic Review 106 (5): 31–34.https://doi.org/10.1257/aer.p20161003.

———. 2018. “Should We Reject the Natural Rate Hypothesis?” Journal of Economic Perspectives 32 (1):97–120. https://doi.org/10.1257/jep.32.1.97.

Blanchard, Olivier, and Francesco Giavazzi. 2002. “Current Account Deficits in the Euro Area: TheEnd of the Feldstein-Horioka Puzzle?” Brookings Papers on Economic Activity 2002 (2): 147–86. http://www.jstor.org/stable/1209205.

Blanchard, Olivier, and Lawrence F. Katz. 1992. “Regional Evolutions.” Brookings Papers on EconomicActivity 1992 (1): 1–75. https://doi.org/10.2307/2534556.

Blanchard, Olivier, and Lawrence H. Summers. 2017. “Rethinking Stabilization Policy: Evolution orRevolution?” Working Paper 24179. National Bureau of Economic Research. https://doi.org/10.3386/w24179.

Fisher, Irving. 1926. “A Statistical Relation Between Unemployment and Price Changes.” InternationalLabour Review 13: 785. https://heinonline.org/HOL/Page?handle=hein.journals/intlr13&id=789&div=&collection=journals.

Forder, James. 2014. Macroeconomics and the Phillips Curve Myth. Oxford University Press. http://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199683659.001.0001/acprof-9780199683659.

Friedman, Milton. 1968. “The Role of Monetary Policy.” The American Economic Review 58 (1): 1–17.http://www.jstor.org/stable/1831652.

Geerolf, François. 2013a. Bubbles and Asset Supply. Ph.D. Thesis Sciences Po.

———. 2013b. “Reassessing Dynamic Efficiency.” UCLA Working Paper. https://fgeerolf.com/r-g.pdf.

———. 2019. “A Theory of Demand Side Secular Stagnation.” UCLA Working Paper. https://fgeerolf.com/hansen.pdf.

Goodhart, Charles. 2018. “Book Review: Myths and Macro: Macroeconomics and the Phillips CurveMyth by James Forder.” Financial World, April, 56–57. https://www.libf.ac.uk/news-and-insights/journal-financial-world.

Hall, Robert E. 2011. “The Long Slump.” American Economic Review 101 (2): 431–69. https://doi.org/10.1257/aer.101.2.431.

Hansen, Alvin H. 1939. “Economic Progress and Declining Population Growth.” The American EconomicReview 29 (1): 1–15. http://www.jstor.org/stable/1806983.

Hooper, Peter, Frederic S Mishkin, and Amir Sufi. 2019. “Prospects for Inflation in a High Pressure Economy:Is the Phillips Curve Dead or Is It Just Hibernating?” Working Paper 25792. National Bureau of EconomicResearch. https://doi.org/10.3386/w25792.

Ilzetzki, Ethan, Carmen M. Reinhart, and Kenneth S. Rogoff. 2019. “Exchange Arrangements Entering theTwenty-First Century: Which Anchor Will Hold?” The Quarterly Journal of Economics 134 (2): 599–646.https://doi.org/10.1093/qje/qjy033.

29

https://doi.org/10.1016/0022-1996(93)90021-O

https://doi.org/10.1016/0022-1996(93)90021-O

https://doi.org/10.3386/w21956

https://doi.org/10.1257/aer.p20161003

https://doi.org/10.1257/jep.32.1.97

http://www.jstor.org/stable/1209205


https://doi.org/10.2307/2534556


https://heinonline.org/HOL/Page?handle=hein.journals/intlr13&id=789&div=&collection=journals

https://heinonline.org/HOL/Page?handle=hein.journals/intlr13&id=789&div=&collection=journals

http://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199683659.001.0001/acprof-9780199683659

http://www.oxfordscholarship.com/view/10.1093/acprof:oso/9780199683659.001.0001/acprof-9780199683659


https://fgeerolf.com/r-g.pdf

https://fgeerolf.com/hansen.pdf

https://fgeerolf.com/hansen.pdf

https://www.libf.ac.uk/news-and-insights/journal-financial-world

https://www.libf.ac.uk/news-and-insights/journal-financial-world

https://doi.org/10.1257/aer.101.2.431

https://doi.org/10.1257/aer.101.2.431



https://doi.org/10.1093/qje/qjy033

Jones, Charles I. 2017. Macroeconomics. W. W. Norton, Incorporated.

Kalantzis, Yannick. 2015. “Financial Fragility in Small Open Economies: Firm Balance Sheets and the SectoralStructure.” The Review of Economic Studies 82 (3): 1194–1222. https://doi.org/10.1093/restud/rdv011.

Kennedy, John Fitzgerald. 1963. Economic Report of the President. Washington, DC: United StatesGovernment Printing Office. https://fraser.stlouisfed.org/files/docs/publications/ERP/1963/ERP_1963.pdf.

Keynes, John Maynard. 1936. The General Theory of Employment, Interest, and Money.

Krugman, Paul. 2018. “Good Enough for Government Work? Macroeconomics Since the Crisis.” OxfordReview of Economic Policy 34 (1-2): 156–68. https://doi.org/10.1093/oxrep/grx052.

Leijonhufvud, Axel. 1967. “Keynes and the Keynesians: A Suggested Interpretation.” The AmericanEconomic Review 57 (2): 401–10. https://www.jstor.org/stable/1821641.

Lucas, Robert E. 1973. “Some International Evidence on Output-Inflation Tradeoffs.” The AmericanEconomic Review 63 (3): 326–34. http://www.jstor.org/stable/1914364.

———. 1976. “Econometric Policy Evaluation: A Critique.” Carnegie-Rochester Conference Series on PublicPolicy 1 (Supplement C): 19–46. https://doi.org/10.1016/S0167-2231(76)80003-6.

Lucas, Robert E., and Thomas J. Sargent. 1979. “After Keynesian Macroeconomics.” Federal Reserve BankMinneapolis Quarterly Review, no. Spr. https://ideas.repec.org/a/fip/fedmqr/y1979isprnv.3no.2.html.

Mankiw, N. Gregory. 2015. Macroeconomics. Macmillan Higher Education.

Martin, Philippe, and Thomas Philippon. 2017. “Inspecting the Mechanism: Leverage and the Great Recessionin the Eurozone.” American Economic Review 107 (7): 1904–37. https://doi.org/10.1257/aer.20150630.

McLeay, Michael, and Silvana Tenreyro. 2019. “Optimal Inflation and the Identification of the Phillips Curve.”NBER Macroeconomics Annual 2019, Volume 34, June. https://www.nber.org/chapters/c14245.

Moretti, Enrico. 2013. “Real Wage Inequality.” American Economic Journal: Applied Economics 5 (1):65–103. https://doi.org/10.1257/app.5.1.65.

Nakamura, Emi, and Jón Steinsson. 2018. “Identification in Macroeconomics.” Journal of EconomicPerspectives 32 (3): 59–86. https://doi.org/10.1257/jep.32.3.59.

Nakamura, Emi, Jón Steinsson, Patrick Sun, and Daniel Villar. 2018. “The Elusive Costs of Inflation:Price Dispersion During the U.S. Great Inflation.” The Quarterly Journal of Economics 133 (4): 1933–80.https://doi.org/10.1093/qje/qjy017.

Phelps, Edmund S. 1967. “Phillips Curves, Expectations of Inflation and Optimal Unemployment over Time.”Economica 34 (135): 254–81. https://doi.org/10.2307/2552025.

Phillips, A. W. 1958. “The Relation Between Unemployment and the Rate of Change of Money Wage Ratesin the United Kingdom, 1861-1957.” Economica 25 (100): 283–99. https://doi.org/10.1111/j.1468-0335.1958.tb00003.x.

Rebelo, Sergio, and Carlos A. Vegh. 1995. “Real Effects of Exchange-Rate-Based Stabilization: An Analysisof Competing Theories.” NBER Macroeconomic Annual, January, 125–88. http://www.nber.org/chapters/c11018.

Robinson, Joan. 1974. “What Has Become of the Keynesian Revolution?” Challenge 16 (6): 6–11.https://fgeerolf.com/phillips/bib/Robinson1974.pdf.

Rodrik, Dani. 2008. “The Real Exchange Rate and Economic Growth.” Brookings Papers on EconomicActivity, no. 2: 365–412. https://doi.org/10.1353/eca.0.0020.

Romer, Christina D., and David H. Romer. 2004. “A New Measure of Monetary Shocks: Derivation andImplications.” American Economic Review 94 (4): 1055–84. https://doi.org/10.1257/0002828042002651.

30

https://doi.org/10.1093/restud/rdv011

https://fraser.stlouisfed.org/files/docs/publications/ERP/1963/ERP_1963.pdf

https://doi.org/10.1093/oxrep/grx052

https://www.jstor.org/stable/1821641


https://doi.org/10.1016/S0167-2231(76)80003-6

https://ideas.repec.org/a/fip/fedmqr/y1979isprnv.3no.2.html

https://doi.org/10.1257/aer.20150630

https://www.nber.org/chapters/c14245

https://doi.org/10.1257/app.5.1.65

https://doi.org/10.1257/jep.32.3.59

https://doi.org/10.1093/qje/qjy017

https://doi.org/10.2307/2552025

https://doi.org/10.1111/j.1468-0335.1958.tb00003.x

https://doi.org/10.1111/j.1468-0335.1958.tb00003.x

http://www.nber.org/chapters/c11018

http://www.nber.org/chapters/c11018

https://fgeerolf.com/phillips/bib/Robinson1974.pdf

https://doi.org/10.1353/eca.0.0020

https://doi.org/10.1257/0002828042002651

———. 2010. “The Macroeconomic Effects of Tax Changes: Estimates Based on a New Measure of FiscalShocks.” American Economic Review 100 (3): 763–801. https://doi.org/10.1257/aer.100.3.763.

Samuelson, Paul A., and Robert M. Solow. 1960. “Analytical Aspects of Anti-Inflation Policy.” The AmericanEconomic Review 50 (2): 177–94. http://www.jstor.org/stable/1815021.

Schmitt-Grohé, Stephanie, and Martin Uribe. 2016. International Macroeconomics.

Sleeman, A. G. 2011. “Retrospectives: The Phillips Curve: A Rushed Job?” Journal of Economic Perspectives25 (1): 223–38. https://doi.org/10.1257/jep.25.1.223.

Smets, Frank, and Rafael Wouters. 2007. “Shocks and Frictions in US Business Cycles: A Bayesian DSGEApproach.” American Economic Review 97 (3): 586–606. https://doi.org/10.1257/aer.97.3.586.

Steil, Benn. 2013. The Battle of Bretton Woods: John Maynard Keynes, Harry Dexter White, and the Makingof a New World Order. Princeton University Press.

Summers, Lawrence H. 1991. “Should Keynesian Economics Dispense with the Phillips Curve?” In Issues inContemporary Economics, 3–20. International Economic Association Series. Palgrave Macmillan, London.

———. 2013. “Conference Address.” IMF Fourtheenth Annual Research Conference in Honor of StanleyFischer. https://www.youtube.com/watch?v=KYpVzBbQIX0.

———. 2017. “Crises in Economic Thought, Secular Stagnation, and Future Economic Research.” NBERMacroeconomics Annual 31 (1): 557–77. https://doi.org/10.1086/690245.

Vegh, Carlos A. 2013. Open Economy Macroeconomics in Developing Countries. MIT Press.

Yellen, Janet. 2017. “Speech by Chair Yellen on the U.S. Economy and Monetary Policy.” https://www.federalreserve.gov/newsevents/speech/yellen20171015a.htm.

Zidar, Owen. 2019. “Tax Cuts for Whom? Heterogeneous Effects of Income Tax Changes on Growth andEmployment.” Journal of Political Economy 127 (3): 1437–72. https://doi.org/10.1086/701424.

31

https://doi.org/10.1257/aer.100.3.763


https://doi.org/10.1257/jep.25.1.223

https://doi.org/10.1257/aer.97.3.586

https://www.youtube.com/watch?v=KYpVzBbQIX0

https://doi.org/10.1086/690245

https://www.federalreserve.gov/newsevents/speech/yellen20171015a.htm

https://www.federalreserve.gov/newsevents/speech/yellen20171015a.htm

https://doi.org/10.1086/701424

A Price Levels and Real Exchange RatesIn this section, I remind the reader of basic concepts in international economics concerning the relationshipbetween price levels, the nominal and real exchange rates. The real exchange rate ε is given as a function ofthe nominal exchange rate E, the home price level P and the foreign price level P ∗ by:

ε = E · PP ∗

Throughout, I assume that the foreign price is fixed, either because I consider a small open economy, orbecause I implicitely neglect the effects of the home economy on the foreign economy. Therefore, shocks tothe home economy are considered, taking other economies as given:

P ∗ = P ∗.

I now relate price levels and real exchange rates, and the first difference of their logarithms, inflation and realexchange rate growth, under different monetary policy arrangements. I consider fixed exchange rates, thenflexible exchange rates with perfect price level or inflation taregeting, and finally a mixed regime, whereby ingeneral both the nominal exchange rate and the price level adjust following a shock to the real exchange rate.

A.1 Fixed Exchange RatesUnder fixed exchange rates (E = E), a higher price level implies a higher level of the real exchange rate:

E = E ⇒ ε =(E

P ∗

)· P.

From the above expression, real exchange rate growth is simply equal to inflation in fixed exchange rates:

∆ log ε = ∆ logP.

As a consequence, a correlation between inflation and unemployment is also a correlation between realexchange rate growth and unemployment. Anticipating on the results, this paper shows that in fact, PhillipsCurves have only robustly been documented under fixed exchange rate regimes.

A.2 Flexible Exchange Rates with Price-level / Inflation targetingUnder flexible exchange rates, and assuming price level targeting P = P , the level of the real exchange ratedepends linearly on the level of the nominal exchange rate:

P = P ⇒ ε =(P

P ∗

)· E.

With flexible exchange rate, under inflation targeting, real exchange rate growth is equal to nominal exchangerate growth:

∆ log ε = ∆ logE.

This is a fortiori the case for price level targeting, as the inflation rate is then also constant and equal to zero.

A.3 Flexible Exchange RatesIf neither the price level nor the nominal exchange rate are stabilized, then the level of the real exchange ratedepends on both:

ε =(

1P ∗

)· E · P.

As a consequence, an rise in the real exchange rate can occur through both through an appreciation of thenominal exchange rate, or a rise in the level of prices.

32

Conditional on real exchange rate growth, the split between inflation and nominal exchange rate growth isundetermined:

∆ log ε︸︷︷︸Real ER Growth

= ∆ logE︸︷︷︸Nominal ER Growth

+ ∆ logP︸︷︷︸inflation

.

A.4 SummaryGiven the definitions above, a relationship between inflation and unemployment in a fixed exchange rateregime, as that considered in Phillips (1958) and Samuelson and Solow (1960), may as well be a relationbetween real exchange rates and unemployment.

33

B Phillips Curve CorrelationsB.1 Cross-Country EvidenceB.1.1 Original Price Phillips Curves

B.1.1.1 Harmonised headline inflation

Table 16: Original Phillips Curve: Harmonised headline inflation and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.22*** -7.7 14.7% 334Crawling Peg -0.06 -0.4 -2.5% 33Crawling Band -0.39** -2.2 11.6% 27

Floating 0.7** 3.2 47.7% 9

Table 17: Original Phillips Curve: Harmonised headline inflation and Unemployment (Quar-terly)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.2*** -5.9 10.4% 291Crawling Peg -0.1 -0.6 -2.3% 25Crawling Band 0.37* 1.8 11% 17

Floating 0.77** 3.2 48.7% 9

B.1.1.2 Core inflation

Table 18: Original Phillips Curve: Core inflation and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.26 -0.7 -1% 43Crawling Peg -0.42 -1.4 0.7% 137Crawling Band -1.05*** -4.7 5.9% 339

Floating -0.13 -0.9 -0.2% 110

Table 19: Original Phillips Curve: Core inflation and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.43 -1.4 2.1% 42Crawling Peg -0.25 -0.8 -0.2% 137Crawling Band -0.61*** -3.3 3.6% 272

Floating -0.12 -0.8 -0.3% 110

34

B.1.1.3 Harmonised core inflation

Table 20: Original Phillips Curve: Harmonised core inflation and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.23*** -9.2 19.9% 334Crawling Peg -0.17 -1.0 -0.2% 32Crawling Band -0.16 -0.8 -2% 21

Floating 0.38** 3.2 47.9% 9

Table 21: Original Phillips Curve: Harmonised core inflation and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.19*** -6.4 11.9% 291Crawling Peg -0.19 -1.2 1.7% 24Crawling Band -0.62 -1.3 5.9% 11

Floating 0.38** 2.5 33.9% 9

B.1.1.4 Consumption deflator growth

Table 22: Original Phillips Curve: Consumption deflator growth and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -0.16 -1.1 0.1% 129

Table 23: Original Phillips Curve: Consumption deflator growth and Unemployment (Quar-terly)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.3*** -5.7 6.3% 466Crawling Peg -0.53*** -3.6 4.5% 259Crawling Band -0.58*** -3.2 2.8% 311

Floating -0.13 -0.8 -0.2% 129

35

B.1.1.5 GDP Deflator growth

Table 24: GDP Deflator growth and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -0.31** -2.1 2.5% 129

Table 25: GDP Deflator growth and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.46*** -4.5 4% 466Crawling Peg -0.71*** -4.1 5.8% 259Crawling Band -1.03*** -4.6 6.1% 311

Floating -0.16 -0.9 -0.1% 129

B.1.1.6 Headline inflation

Table 26: Original Phillips Curve: Headline inflation and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.05 -0.2 -1.2% 79Crawling Peg -0.28 -1.0 0% 171Crawling Band -0.51*** -2.8 2.3% 277

Floating -0.21 -1.3 0.5% 114

36

B.1.2 Accelerationist Phillips Curves

B.1.2.1 Harmonised headline inflation

Table 27: Accelerationist Phillips Curve: Harmonised headline inflation and Unemployment(Annual)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.09*** -3.2 2.8% 326Crawling Peg -0.09 -0.6 -2.6% 26Crawling Band -0.73*** -5.0 47.5% 25

Floating 0.07 0.2 -11.7% 8

Table 28: Accelerationist Phillips Curve: Harmonised headline inflation and Unemployment(Quarterly)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.13*** -3.6 4.1% 284Crawling Peg 0.03 0.1 -5.2% 19Crawling Band -0.33** -2.6 24.9% 16

Floating 0.09 0.4 -10.6% 8

B.1.2.2 Core inflation

Table 29: Accelerationist Phillips Curve: Core inflation and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.25 -0.9 -0.4% 39Crawling Peg -0.26 -1.6 1.2% 132Crawling Band -0.07 -0.5 -0.2% 329

Floating -0.13** -2.3 3.6% 109

Table 30: Accelerationist Phillips Curve: Core inflation and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.12 -0.5 -2.1% 38Crawling Peg -0.15 -0.7 -0.4% 132Crawling Band -0.13 -1.1 0% 265

Floating -0.11 -1.5 1.1% 109

37

B.1.2.3 Harmonised core inflation

Table 31: Accelerationist Phillips Curve: Harmonised core inflation and Unemployment(Annual)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.07*** -3.5 3.3% 326Crawling Peg -0.06 -0.4 -3.2% 25Crawling Band -0.67*** -7.4 72.7% 19

Floating 0.09 0.6 -7.8% 8

Table 32: Accelerationist Phillips Curve: Harmonised core inflation and Unemployment(Quarterly)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.1*** -4.1 5.3% 284Crawling Peg 0 0.0 -5.5% 18Crawling Band -1.04** -3.1 44.1% 10

Floating 0.08 0.4 -10.3% 8

B.1.2.4 Consumption deflator growth

Table 33: Accelerationist Phillips Curve: Consumption deflator growth and Unemployment(Annual)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.07** -2.2 0.8% 531Crawling Peg -0.24*** -2.8 1.9% 341Crawling Band -0.22* -1.8 0.5% 441

Floating -0.21*** -2.7 4.6% 128

Table 34: Accelerationist Phillips Curve: Consumption deflator growth and Unemployment(Quarterly)

Exchange Rate Regime PC Slope t-stat Adj. R2 NFixed / Peg -0.1* -1.8 0.5% 450Crawling Peg -0.19** -2.0 1.1% 252Crawling Band -0.08 -0.4 -0.3% 308

Floating -0.21** -2.4 3.5% 128

38

B.1.2.5 GDP Deflator, growth

Table 35: GDP Deflator growth and Unemployment (Annual)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.04 -1.2 0.1% 528Crawling Peg -0.14* -1.9 0.8% 336Crawling Band -0.15 -1.3 0.2% 443

Floating -0.16* -1.9 1.9% 128

Table 36: GDP Deflator growth and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.1 -0.7 -0.1% 450Crawling Peg -0.03 -0.1 -0.4% 252Crawling Band -0.1 -0.3 -0.3% 308

Floating -0.04 -0.3 -0.7% 128

B.1.2.6 Headline inflation

Table 37: Accelerationist Phillips Curve: Headline inflation and Unemployment (Quarterly)Exchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.21 -1.2 0.5% 73Crawling Peg -0.21 -1.2 0.2% 165Crawling Band -0.08 -0.6 -0.2% 274

Floating -0.32*** -2.7 5.3% 114

39

B.1.3 House Prices Phillips Curves

B.1.3.1 House Prices (BIS)

Table 38: House Prices (BIS) and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -3.26*** -5.2 16.7% 127

B.1.3.2 House Prices (Dallas Fed)

Table 39: Dallas Fed Nominal House Price Growth and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -1.31*** -3.8 9.7% 125

Table 40: Dallas Fed Real House Price Growth and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.95*** -7.3 16.9% 258Crawling Peg -0.59** -2.2 2.4% 156Crawling Band -0.41* -1.9 1.2% 209

Floating -1.28*** -4.2 11.7% 125

B.1.3.3 House Prices (OECD)

Table 41: OECD House Price Growth and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N


Floating -1.35*** -4.0 10.2% 128

Table 42: OECD Real House Price Growth and UnemploymentExchange Rate Regime PC Slope t-stat Adj. R2 N

Fixed / Peg -0.77*** -6.0 9.7% 326Crawling Peg -0.66*** -3.0 3.3% 237Crawling Band -0.69*** -3.1 3% 277

Floating -1.24*** -4.1 11% 128

40

B.1.4 CPI Components


CPI: 01 - Food & non-Alcoholic beverages -0.62*** -1.73*** -2.19*** -0.21CPI: 01-12 - All items -0.5*** -1.12*** -1.13*** -0.11

CPI: 02 - Alcoholic beverages, tobacco & narcotics 0.07 -0.51 0.5 0.44CPI: 03 - Clothing & footwear -0.47*** 0.39 0.11 -2.15***

CPI: 04 - Housing, water, electricity, gas & other fuels -0.48*** -0.26 -0.82*** -1.56***CPI: 04.1 - CPI Actual rentals for housing -1.55*** -0.27 -0.26 -0.95***CPI: 04.2 - CPI Imputed rentals for housing 0.23 -1.25*** -1.22*** -2.04***

CPI: 04.3 - CPI Maintenance & repairs of dwellings -0.88*** -0.24 -0.28 -0.47CPI: 04.4 - CPI Water supply & Misc. services -0.37*** 1.57** 0.54** -0.27CPI: 04.5 - CPI Electricity, gas & other fuels 0.12 0.96 -0.48 -1.18*CPI: 05 - Furnishings, household equipment -0.57*** -0.06 -0.02 -1.83***

CPI: 06 - Health -0.3*** 0.09 0.19 -0.04CPI: 07 - Transport -0.25** 0.74* 0.16 1.84***

CPI: 07.2.2 - CPI Fuels & lubricants -0.21 0.27 0.2 1.37CPI: 08 - Communication 0.14 -0.07 0.84* 0.16

CPI: 09 - Recreation & culture -0.43*** 0.01 -0.02 -2.32***CPI: 10 - Education -0.17 0.29 0.38 -0.7

CPI: 11 - Restaurants & hotels -0.63*** -0.5 -0.84*** -1.65***CPI: 12 - Miscellaneous goods & services -0.39*** 0.16 -0.21 -0.53

CPI: All items non-food non-energy -0.58*** -0.91*** -1.03*** 0.1CPI: Energy -0.12 -0.41 -0.79** 0CPI: Goods -0.35*** 0.36 0 -0.9*CPI: Housing -0.73*** -1.21*** -0.72** -0.88***

CPI: Housing excl. imp. rent -1.07*** 0.05 -1.5*** -0.99***CPI: Services -0.69*** -0.61* 0.58** -1.03***

CPI: Services less housing -0.64*** -0.09 -0.63* -0.99***CPI: Services less housing (Housing excl. imp. rent) -0.52*** -0.99*** -4.02*** 0.17

HICP: 01 - Food & non-Alcoholic beverages -0.45*** 0.33 -1.46** 0.95*HICP: 02 - Alcoholic beverages, tobacco & narcotics 0.08 0.67** -1.27** 2.03***

HICP: 03 - Clothing & footwear -0.4*** 0.36 -0.36 0.5HICP: 04 - Housing, water, electricity, gas & other fuels -0.27*** 2.59*** -1.44** -0.6

HICP: 04.1 - HICP Actual rentals for housing -1.42*** 2.53*** -1.78** 0.06HICP: 04.3. - HICP Maintenance & repairs of dwellings -0.82*** 0.81*** -1.77*** 3.06***

HICP: 04.4 - HICP Water supply & Misc. services -0.44** 2.47*** -1.19* 2.12**HICP: 04.5 - HICP Electricity, gas & other fuels 0.09 3.25*** -1.34* 4.33HICP: 05 - Furnishings, household equipment -0.46*** 0.51** -1.01 0.5

HICP: 06 - Health -0.21** 0.34 -1.73** 0.18HICP: 07 - Transport -0.23** 2.06*** -1.2* 1.09

HICP: 07.2.2 - CPI Fuels & lubricants -0.2 3.4*** -1.33 10.59***HICP: 08 - Communication 0.35*** 2.06*** -0.57 0.91*

HICP: 09 - Recreation & culture -0.39*** 0.62** -0.89 0.17HICP: 10 - Education -0.15 0.71 -1.58** -0.14

HICP: 11 - Restaurants & hotels -0.59*** 0.67** -1.9*** -0.17HICP: 12 - Miscellaneous goods & services -0.38*** 0.8*** -1.55*** 0.14

HICP: All items -0.32*** 0.55** -1.01* 0.27HICP: Energy -0.09 3.45*** -1.37** 7.07***

HICP: Ind. excl. energy, food, alcohol & tobacco -0.37*** 0.57* -1.05* 0.66**HICP: Services -0.47*** 1.24*** -1.57** 0.81***

41


CPI: 01 - Food & non-Alcoholic beverages -0.77*** -3.04*** -3.07*** 0.04CPI: 01-12 - All items -0.63*** -1.97** -1.46*** 0.15

CPI: 02 - Alcoholic beverages, tobacco & narcotics 0.3** -0.83* 0.94** 0.6CPI: 03 - Clothing & footwear -0.67*** 0.89 0.31 -3.69***

CPI: 04 - Housing, water, electricity, gas & other fuels -0.4** -0.03 -0.77** -1.82***CPI: 04.1 - CPI Actual rentals for housing -1.99*** -0.21 0.21 -1.34**CPI: 04.2 - CPI Imputed rentals for housing 0.84 -1.37** -1.12** -3.51***

CPI: 04.3 - CPI Maintenance & repairs of dwellings -1.15*** 0.11 -0.13 -0.48CPI: 04.4 - CPI Water supply & Misc. services -0.3* 2.3** 1.07*** -0.45CPI: 04.5 - CPI Electricity, gas & other fuels 0.62*** 1.86** -0.34 -2.12**CPI: 05 - Furnishings, household equipment -0.74*** 0.11 0.25 -2.83***

CPI: 06 - Health -0.39*** 0.25 0.52* -0.42CPI: 07 - Transport -0.27* 1.09* 0.58 2.02***

CPI: 07.2.2 - CPI Fuels & lubricants -0.05 0.34 0.37 1.55CPI: 08 - Communication 0.26 0.41 1.45** -0.03

CPI: 09 - Recreation & culture -0.59*** 0.13 0.29 -3.64***CPI: 10 - Education -0.16 0.14 0.86** -1

CPI: 11 - Restaurants & hotels -0.77*** -0.73 -0.77** -2.54***CPI: 12 - Miscellaneous goods & services -0.46*** 0.13 -0.01 -0.66

CPI: All items non-food non-energy -0.74*** -1.07*** -1.28*** 0.36CPI: Energy 0.15 -0.2 -0.94** 0.72CPI: Goods -0.32** 0.85 0.23 -1.97***CPI: Housing -0.99*** -1.48*** -0.65 -1.1***

CPI: Housing excl. imp. rent -1.39*** 0.26 -1.96** -1.26**CPI: Services -0.81*** -0.51 1.39*** -1.24**

CPI: Services less housing -0.77*** 0.09 -0.89* -1.17***CPI: Services less housing (Housing excl. imp. rent) -0.55*** -0.96* -5.91*** 0.13

HICP: 01 - Food & non-Alcoholic beverages -0.47*** 0.66** -1.31* 1.66**HICP: 02 - Alcoholic beverages, tobacco & narcotics 0.34*** 0.98*** -1.05 3.03***

HICP: 03 - Clothing & footwear -0.55*** 0.83** 0.43 0.29HICP: 04 - Housing, water, electricity, gas & other fuels -0.09 3.96*** -1.11 -0.72

HICP: 04.1 - HICP Actual rentals for housing -1.82*** 3.52*** -1.18 -0.16HICP: 04.3. - HICP Maintenance & repairs of dwellings -1.02*** 1.34*** -1.5 5.13***

HICP: 04.4 - HICP Water supply & Misc. services -0.41* 3.22*** -0.47 2.03*HICP: 04.5 - HICP Electricity, gas & other fuels 0.58** 4.91*** -1.09 5.53*HICP: 05 - Furnishings, household equipment -0.55*** 1.06*** -0.62 0.84**

HICP: 06 - Health -0.24** 0.57 -1.77 0.34*HICP: 07 - Transport -0.16 2.71*** -0.91 1.46

HICP: 07.2.2 - CPI Fuels & lubricants 0 4.1*** -1.08 14.97***HICP: 08 - Communication 0.63*** 3.53*** -0.24 1.34*

HICP: 09 - Recreation & culture -0.49*** 0.99*** -0.53 0.17HICP: 10 - Education -0.05 0.46 -1.43 -0.3

HICP: 11 - Restaurants & hotels -0.69*** 1*** -1.85** -0.06HICP: 12 - Miscellaneous goods & services -0.4*** 1.21*** -1.37 0.25

HICP: All items -0.32*** 1.11*** -0.74 0.39HICP: Energy 0.27 4.8*** -1.13 9.75***

HICP: Ind. excl. energy, food, alcohol & tobacco -0.43*** 1.26*** -0.65 0.78**HICP: Services -0.51*** 1.92*** -1.3 1.04***

42


CPI: 01 - Food & non-Alcoholic beverages -0.35*** -0.7*** -1.16** -0.24CPI: 01-12 - All items -0.28*** -0.46*** -0.73*** -0.2

CPI: 02 - Alcoholic beverages, tobacco & narcotics -0.06 -0.15 0.04 0.14CPI: 03 - Clothing & footwear -0.23*** 0.12 0.07 -0.94***

CPI: 04 - Housing, water, electricity, gas & other fuels -0.34*** -0.33 -0.55*** -0.93***CPI: 04.1 - CPI Actual rentals for housing -0.77*** -0.24* -0.41*** -0.49***CPI: 04.2 - CPI Imputed rentals for housing -0.01 -0.85*** -0.77*** -0.94***

CPI: 04.3 - CPI Maintenance & repairs of dwellings -0.47*** -0.29 -0.34*** -0.32**CPI: 04.4 - CPI Water supply & Misc. services -0.24*** 0.63 0.03 -0.16CPI: 04.5 - CPI Electricity, gas & other fuels -0.06 0.17 -0.4** -0.53CPI: 05 - Furnishings, household equipment -0.32*** -0.13 -0.07 -0.88***

CPI: 06 - Health -0.18*** -0.08 0.06 0CPI: 07 - Transport -0.15** 0.25 -0.02 1.03***

CPI: 07.2.2 - CPI Fuels & lubricants -0.18 -0.05 -0.02 0.63CPI: 08 - Communication 0.03 -0.19 0.4* 0.16

CPI: 09 - Recreation & culture -0.22*** -0.05 -0.07 -1.12***CPI: 10 - Education -0.13* 0.11 0.08 -0.4*

CPI: 11 - Restaurants & hotels -0.37*** -0.32* -0.54*** -0.81***CPI: 12 - Miscellaneous goods & services -0.24*** 0.07 -0.22* -0.39**

CPI: All items non-food non-energy -0.33*** -0.56*** -0.59*** -0.06CPI: Energy -0.14 -0.42** -0.56*** -0.42CPI: Goods -0.23*** 0.04 -0.08 -0.26CPI: Housing -0.39*** -0.73*** -0.53*** -0.49***

CPI: Housing excl. imp. rent -0.56*** -0.09 -0.83*** -0.53***CPI: Services -0.42*** -0.47*** 0.06 -0.61***

CPI: Services less housing -0.34*** -0.19 -0.46** -0.6***CPI: Services less housing (Housing excl. imp. rent) -0.33*** -0.67*** -2.16*** 0.07

HICP: 01 - Food & non-Alcoholic beverages -0.27*** 0.06 -0.91*** 0.36HICP: 02 - Alcoholic beverages, tobacco & narcotics -0.07 0.33* -0.82** 0.92***

HICP: 03 - Clothing & footwear -0.2*** 0.04 -0.27 0.37HICP: 04 - Housing, water, electricity, gas & other fuels -0.23*** 1.04*** -0.94*** -0.38

HICP: 04.1 - HICP Actual rentals for housing -0.69*** 1.14*** -1.16*** 0.17HICP: 04.3. - HICP Maintenance & repairs of dwellings -0.44*** 0.28** -0.98*** 1.22**

HICP: 04.4 - HICP Water supply & Misc. services -0.34** 1.14*** -0.87** 1.34**HICP: 04.5 - HICP Electricity, gas & other fuels -0.1 1.38*** -0.91** 2.4*HICP: 05 - Furnishings, household equipment -0.27*** 0.1 -0.6** 0.19

HICP: 06 - Health -0.15*** 0.01 -0.87** 0.06HICP: 07 - Transport -0.15** 0.9*** -0.67** 0.47

HICP: 07.2.2 - CPI Fuels & lubricants -0.15 1.54*** -0.68 4.51*HICP: 08 - Communication 0.12* 0.81** -0.13 0.36

HICP: 09 - Recreation & culture -0.22*** 0.25 -0.53* 0.07HICP: 10 - Education -0.15** 0.47 -0.87** 0.06

HICP: 11 - Restaurants & hotels -0.34*** 0.26* -1.09*** -0.16*HICP: 12 - Miscellaneous goods & services -0.25*** 0.31** -0.87*** 0.02

HICP: All items -0.21*** 0.15 -0.64** 0.1HICP: Energy -0.17 1.53*** -0.87** 3.28**

HICP: Ind. excl. energy, food, alcohol & tobacco -0.23*** 0.05 -0.63** 0.37**HICP: Services -0.29*** 0.45** -0.88*** 0.45***

43

B.2 United Kingdom

1861

1862

1863

1864

1865

1866

1867

18681869

1870

1871

1872

1873

1874

1875

18761877

18781879

18801881

18821883

1884

1885

1886

1887

1888

1889

1890

1891

1892

1893

1894

1895

1896189718981899

1900

1901

19021903 1904

19051906

1907

1908

1909

19101911

1912

1913

−2%

−1%

0%

1%

2%

3%

4%

5%

3% 4% 5% 6% 7% 8%Unemployment Rate

Wag

e In

flatio

n

Figure 18: U.K. Phillips Curve (1860-1915): Unemployment and Wage Inflation.

3.5

4.0

4.5

5.0

5.5

6.0

14 15 16 17 18 19 20 21 22 23 24 25 26 27

Figure 19: U.K. Dollar / Pound Exchange Rate (1914-1927)

44

−20%

−15%

−10%

−5%

0%

5%

10%

15%

20%

25%

1860 1870 1880 1890 1900 1910 1920 1930 1940 1950

Unemployment Rate

Wage Inflation

Figure 20: U.K. Phillips Curve (1860-1950): Unemployment and Wage Inflation.

B.3 United States

Missing Deflation Missing Inflation−2%

−1%

0%

1%

2%

3%

4%

5%

6%

05 06 07 08 09 10 11 12 13 14 15 16 17 18 19

Rol

ling

2−Ye

ar C

PI I

nfla

tion

Consumer Price Index for All Urban Consumers: All items less shelterConsumer Price Index for All Urban Consumers: Shelter

Figure 21: U.S. Missing Inflation, Deflation

45

9293

949596

97

9899

00

0102

03

040506

07

08

09

10

11

1213

1415

16

17

18

19

20

21

22

23

24

25

26

27

2829

30

31

3233

34

3536

37 38

39

4041

4243

44

46

47

48

49

50

5152

53 54

5556

5758

5960

−10%

−5%

0%

5%

10%

15%

20%

2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24% 26%Unemployment Rate

Infla

tion

Figure 23: U.S. Phillips Curve (1890-1960): Unemployment and Price Inflation.

StagflationMissing Inflation

Missing DeflationMissing Inflation

−16%

−14%

−12%

−10%

−8%

−6%

−4%

−2%

0%

2%

4%

6%

8%

10%

70 80 90 00 10 20

Real House Price inflation (Shiller's Index)Rent inflation (rel. to overall CPI)Unemployment Rate

Figure 22: U.S. Unemployment and Relative Rent Price

46

1891

1892

1893 18941895

18961897

1898

1899

1900

19011902

190319041905

19061907

1908

1909

1910

1911

1912

1913 19141915

1916

1917

1918

1919

1920

1921

1922

1923

1924

1925

1926

19271928

1929

1930

19311932

−18%

−16%

−14%

−12%

−10%

−8%

−6%

−4%

−2%

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24% 26%Unemployment Rate

Pric

e In

flatio

n

Figure 24: U.S. Price Phillips Curve (1890-1933).

$15

$20

$25

$30

$35

$40

20 23 26 29 3719 21 24 27 33 38

Gol

d P

rices

($/

ounc

e)

Figure 25: June 5, 1933: FDR takes U.S. off the Gold Standard

47

12

13

14

15

16

17

18

22 23 24 25 26 27 28 29 30 31 32 33 34 35

CP

I

Figure 26: U.S. CPI (1922-1935)

1956 1957

19581959

1960

1961

1962

1963

1964

1965

1966

1967

0.0%

0.5%

1.0%

1.5%

2.0%

2.5%

3.0%

3.5%

4.0% 4.5% 5.0% 5.5% 6.0% 6.5%Unemployment Rate

Pric

e In

flatio

n

Figure 27: U.S. Price Phillips Curve (1955-1968).

48

92.5

95.0

97.5

100.0

102.5

105.0

107.5

110.0

22 23 24 25 26 27 28 29 30 31 32 33 34 35

Wag

es

Figure 28: U.S. Index of Manufacturing Wage Rates (07/1922 - 07/1935)

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

22%

24%

26%

29 30 31 32 33 34 35 36 37 38 39 40 41 42

Une

mpl

oym

ent R

ate

Figure 29: U.S. Unemployment (1929-1942)

49

1892

1893

1894

18951896

18971898

1899

1900

1901

19021903

190419051906

19071908

1909

1910

1911

1912

1913

1914 19151916

1917

19181919

1920

1921

1922

1923

1924

1925

1926

1927

19281929

1930

1931

19321933

−15%

−10%

−5%

0%

5%

10%

15%

20%

5% 10% 15% 20% 25%Unemployment Rate

Infla

tion

Figure 30: U.S. Phillips Curve (1890-1933): Unemployment and Price Inflation.

1892

1893

1894

18951896

18971898

1899

1900 19011902

1903

190419051906

1907

1908

1909

1910

1911

1912

1913

1914 19151916

1917

19181919

1920

1921

1922

1923

1924

1925

1926

1927

19281929

1930

1931

19321933

1934

19351936

1937 1938

1939

1940

−12%

−10%

−8%

−6%

−4%

−2%

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

2% 4% 6% 8% 10% 12% 14% 16% 18% 20% 22% 24% 26%Unemployment Gap

Infla

tion

a

a

1890−1933 ($20 Gold Standard)

1933−1941 ($35 Gold Standard)

Figure 31: U.S. Phillips Curve (1960-2019)

50

0%

1%

2%

3%

4%

5%

6%

7%

8%

9%

10%

−2% −1% 0% 1% 2% 3% 4% 5%Unemployment Gap

Infla

tion

1960−1971 (Bretton Woods)

1971−2019 (Flexible ER)


0%

1%

2%

3%

4%

5%

6%

7%

8%

9%

10%


Infla

tion

1960−1971

1971−2019


51

−4%

−3%

−2%

−1%

0%

1%

2%

3%

4%

5%


Cha

nge

in In

flatio

n1960−1971

1971−2019

Figure 34: U.S. Accelerating Phillips Curve (1960-2019)

−4%

−3%

−2%

−1%

0%

1%

2%

3%

4%

5%


Cha

nge

in In

flatio

n

1960−1983

1984−1999

2000−2019

Figure 35: U.S. Flattening Accelerating Phillips Curve? (1960-2019)

52

0%

1%

2%

3%

4%

5%

6%

7%

8%

9%

10%


Infla

tion

1960−1971

1971−1999

2000−2019

Figure 36: U.S. Original Phillips Curve (1960-2019)

53

CPI Components.

Table 46: CPI Components (All)Sector coef

Lodging away from home -0.84**Owners’ equivalent rent of residences -0.25*

Owners’ equivalent rent of primary residence -0.25*Energy services 0.86**

Consumer Price Index - All Urban Consumers - Electricity 0.88***Water and sewer and trash collection services 0.43***

New vehicles 0.8***Used cars and trucks 2.03***

Table 47: Phillips Curves on CPI Components, Before 1971Item (U.S. City Average, All Urban consumers, SA) PC Slope t-stat Adj. R2

Prescription drugs -0.8** -2.9 37.5%Women’s dresses -3.47*** -7.6 82.7%

Furniture and bedding -1.05*** -6.6 78.2%Commodities less food -0.78*** -9.8 36.3%Energy commodities -1.04*** -9.1 34.6%

Commodities less food and energy commodities -0.73*** -8.0 28.9%Fresh whole milk -1.84*** -10.5 28.3%

All items less food and energy -0.8*** -7.8 27.6%Medical care services -0.92*** -7.8 26.2%

Nondurables -0.76*** -7.8 26%All items less energy -0.77*** -7.4 25.5%

Fuel oil -1.46*** -8.6 25.1%Meats, poultry, fish, and eggs -2.77*** -3.6 24.9%

Commodities -0.67*** -7.5 24.8%Men’s and boys’ apparel -1.16*** -9.5 24.4%

All items less food -0.83*** -9.5 24.3%Apparel less footwear -1.3*** -9.4 24.1%

Women’s and girls’ apparel -1.4*** -9.3 23.7%Apparel -1.23*** -9.1 22.8%

All items less medical care -0.7*** -6.8 22.2%All items -0.93*** -8.6 21%

Fuel oil and other fuels -1.45*** -8.5 20.6%Household energy 1.71*** 15.0 86.1%Other food at home 2.57*** 11.7 79.2%

Housekeeping supplies 1.38*** 11.0 77%Medical care commodities 0.97*** 9.7 72.2%

All items less food, shelter, and energy 0.3*** 6.9 56.4%All items less food and shelter 0.35*** 6.7 54.8%

Motor vehicle maintenance and repair 0.7*** 5.1 41.2%Services less energy services 0.86*** 4.8 37.5%

Alcoholic beverages 0.54*** 4.5 35.2%All items less food, shelter, energy, and used cars and trucks 0.23*** 4.4 33.6%

Housing 0.82*** 3.7 26.1%Other goods and services 0.33*** 3.6 24.4%

54

Table 48: Phillips Curves on CPI Components, After 1971Item (U.S. City Average, All Urban consumers, SA) PC Slope t-stat Adj. R2

Other recreation services -0.5*** -15.3 49%Rent of shelter -0.4*** -17.4 47%

Information technology commodities -0.7*** -7.7 37%Club memberships -0.6*** -11.2 34%

Televisions -2.2*** -10.2 28%Moving, storage, freight expense -0.9*** -9.5 27%

Personal care -0.2*** -8.7 24%Admission to movies, theaters, and concerts -0.4*** -8.3 23%

Lodging away from home -0.8*** -8.1 21%Cigarettes -0.4*** -5.1 20%

Transportation commodities less motor fuel 0.6*** 11.1 56%Toys, games, hobbies and playground equipment 1*** 11.8 53%

Parking fees and tolls 1*** 9.3 50%Recreation commodities 0.3*** 8.4 41%

Water and sewerage maintenance 0.6*** 16.2 40%Water and sewer and trash collection services 0.4*** 12.0 37%

Education and communication services 0.4*** 7.4 35%New cars and trucks 0.4*** 11.0 33%New motorcycles 1.2*** 6.8 29%

Women’s underwear, nightwear, sportswear and accessories 0.8*** 9.4 27%College tuition and fees 0.8*** 12.9 26%

Lunchmeats 0.8*** 5.8 25%New trucks 0.7*** 10.7 22%

Inpatient hospital services 0.4*** 8.2 21%Other recreational goods 0.5*** 8.0 20%Medical care commodities 0.8*** 12.1 20%

Tuition, other school fees, and childcare 0.7*** 11.0 20%

55

C Identified MomentsC.1 U.S. Monetary Shocks (Romer and Romer (2004)): Response to a 1%

Increase in the Federal Funds Rate

−1.25%

−1.00%

−0.75%

−0.50%

−0.25%

0.00%

0.25%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 37: Employment

56

−1.5%

−1.0%

−0.5%

0.0%

0.5%

1.0%

1.5%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 38: Real GDP

−1.0%

−0.5%

0.0%

0.5%

1.0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 39: Real Consumption

57

−6%

−5%

−4%

−3%

−2%

−1%

0%

1%

2%

3%

4%

5%

6%

7%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 40: Real Imports

−1%

0%

1%

2%

3%

4%

5%

6%

7%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 41: Unit Labor Costs

58

−6%

−5%

−4%

−3%

−2%

−1%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 42: House Prices

−2%

−1%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 43: Average Nominal Wages

59

0%

1%

2%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 44: Men’s Apparel Prices

C.2 U.S. Fiscal Shocks (Romer and Romer (2010)): Response to a 1% of GDPTax Increase

−2.5%

−2.0%

−1.5%

−1.0%

−0.5%

0.0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 45: Employment

60

−3%

−2%

−1%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 46: Real GDP

−4%

−3%

−2%

−1%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 47: Real Consumption

61

−16%

−14%

−12%

−10%

−8%

−6%

−4%

−2%

0%

2%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 48: Real Imports

−4%

−2%

0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 49: Unit Labor Costs

62

−9%

−8%

−7%

−6%

−5%

−4%

−3%

−2%

−1%

0%

1%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 50: House Prices

−5%

−4%

−3%

−2%

−1%

0%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 51: Average Nominal Wages

63

−2%

−1%

0%

1%

2%

3%

4%

Q0 Q2 Q4 Q6 Q8 Q10 Q12 Q14 Q16 Q18 Q20

Figure 52: Men’s Apparel Prices

D Regional Phillips CurvesD.1 U.S. States

Alabama

Alaska

Arizona

Arkansas

California

Colorado

ConnecticutDelaware

District of Columbia

Florida

Georgia

Hawaii

Idaho

Illinois

Indiana

IowaKansas

Kentucky

Louisiana

Maine

Maryland

Massachusetts

Michigan

Minnesota

Mississippi

Missouri

Montana

Nebraska

Nevada

New HampshireNew Jersey

New Mexico

New York

North CarolinaNorth Dakota

OhioOklahoma

OregonPennsylvania

Rhode Island

South CarolinaSouth DakotaTennessee

TexasUtah

Vermont

Virginia

Washington

West Virginia

Wisconsin

Wyoming

20%

30%

40%

50%

60%

70%

80%

90%

100%

110%

−3% −2% −1% 0% 1% 2% 3%∆ Unemployment Rate 99−06

∆ H

ouse

Pric

es 9

9−06


64

Alabama

Alaska

Arizona

Arkansas

California

Colorado

Connecticut

Delaware

District of Columbia

FloridaGeorgia

Hawaii

Idaho

IllinoisIndiana

Iowa

KansasKentuckyLouisiana

Maine

Maryland

Massachusetts

Michigan

MinnesotaMississippiMissouri

Montana

Nebraska

Nevada

New Hampshire

New Jersey

New Mexico

New York

North Carolina

North DakotaOhioOklahoma

Oregon

Pennsylvania

Rhode Island

South Carolina

South Dakota

TennesseeTexas

Utah

Vermont

Virginia

Washington

West Virginia

Wisconsin

Wyoming

−20%

−10%

0%

10%

20%

30%

40%

50%

−2% −1% 0% 1% 2% 3%∆ Unemployment Rate 90−95

∆ H

ouse

Pric

es 9

0−95


D.2 U.S. MSAs

Brownsville−Harlingen TX

El Centro CA

Fresno CA

Gulfport−Biloxi−Pascagoula MS

Hanford−Corcoran CA

Kokomo INLafayette−West Lafayette IN

Madera CA

McAllen−Edinburg−Mission TX

Ocean City NJ

Odessa TX

Panama City FL

Salinas CA

Visalia−Porterville CA

Yuma AZ

−30%

−20%

−10%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

110%

120%

−12% −10% −8% −6% −4% −2% 0% 2% 4% 6% 8% 10% 12%∆ Unemployment Rate 99−06

∆ H

ouse

Pric

es 9

9−06


65

Bay City MI

Flint MI Hanford−Corcoran CA

Janesville−Beloit WI

Kingston NY

Los Angeles−Long Beach−Anaheim CA

Ocean City NJ

Oxnard−Thousand Oaks−Ventura CA

Saginaw MI

San Luis Obispo−Paso Robles−Arroyo Grande CA

Toledo OH Visalia−Porterville CA

Winchester VA−WV

Yuma AZ

−30%

−20%

−10%

0%

10%

20%

30%

40%

50%

60%

−10% −8% −6% −4% −2% 0% 2% 4% 6% 8% 10% 12%∆ Unemployment Rate 90−95

∆ H

ouse

Pric

es 9

0−95


66

D.3 U.S. Counties

−90%

−80%

−70%

−60%

−50%

−40%

−30%

−20%

−10%

0%

10%

20%

30%

40%

−8% −6% −4% −2% 0% 2% 4% 6% 8% 10% 12% 14%Change in Unemployment Rate 06−09

Cha

nge

in H

ouse

Pric

es 0

6−09

Figure 57: U.S. Regional Phillips Curve (06-09)

D.4 Euro Area Countries

AustriaBelgiumGermany

Spain

Estonia

Finland

France

Greece

Ireland

Italy

Luxembourg

Latvia

Netherlands

Portugal

Slovenia

5%

10%

15%

20%

25%

30%

−8% −6% −4% −2% 0% 2% 4%Change in Unemployment Rate (2001−2007)

Infla

tion

(200

1−20

07)

Figure 58: Euro Area Correlation (2001-2007)

67

E TheoryE.1 Marginal Product of CapitalIn the case where 0 < K/AL ≤ k, the marginal product of capital ∂F/∂K is:

∂F

∂K= αKα−1 (AL)1−α

(1− 1

2kK

AL

)α− α

2kALKα (AL)1−α

(1− 1

2kK

AL

)α−1

= αKα−1 (AL)1−α(

1− 12k

K

AL

)α−1 [(1− 1

2kK

AL

)− 1

2kK

AL

]∂F

∂K= α

(K

AL

)α−1(1− 1

2kK

AL

)α−1(1− 1

k

K

AL

)You can note already that at the maximum level of capital, the marginal product of capital is equal to zero,as then:

MPK = 1− 1k

K

AL= 0

Whenever K/AL > k the marginal product of capital is also equal to zero: there is satiation of capital in theproduction function.

To conclude, the gross Marginal Product of Capital is equal to:

MPK = ∂F

∂K=

α

(K

AL

)α−1(1− 1

2kK

AL

)α−1(1− 1

k

K

AL

)if 0 < K

AL≤ k

0 if K

AL> k

The Marginal Product of Capital can be expressed solely as a function of k ≡ K/AL, as:

MPK = ∂F

∂K=

αkα−1

(1− k

2k

)α−1(1− k

k

)if 0 < k ≤ k

0 if k ≥ k.

Note that we could have obtained this expression by differentiating the intensive form of the productionfunction directly. The intensive form is:

y = f(k) =

kα(

1− 12kk

)αif 0 < k ≤ k(

k

2

)αif k ≥ k

Therefore, when the k is below the threshold value k, we can compute the marginal product of capital givenby f ′(k):

f ′(k) = αkα−1(

1− k

2k

)α− α

2kkα(

1− k

2k

)α−1

= αkα−1(

1− k

2k

)α−1 [(1− k

2k

)− k

2k

]f ′(k) = αkα−1

(1− k

2k

)α−1(1− k

k

)Therefore:

f ′(k) =

αkα−1

(1− k

2k

)α−1(1− k

k

)if 0 < k ≤ k

0 if k ≥ k.

68

E.2 Elasticity of Substitution σ

By definition, the elasticity of substitution between capital and labor is the percentage change in the capitalover labor ratio, when ratio of the price of capital over the wage rate (which is equal to the marginal rate oftransformation) changes by 1%:

σ

(K

L

)= −

d log(K

L

)d log

(∂F/∂K

∂F/∂L

) = −d log

(K

AL

)d log

(∂F/∂K

∂F/∂AL

) .We note that this elasticity is the same regardless of whether it is computed with labor L, or efficiencyunits of labor AL. The latter will prove more convenient. To compute the marginal rate of transformation(∂F/∂K)/(∂F/∂AL), we first compute the marginal product of capital ∂F/∂K with:

Y = F (K,L) =

Kα (AL)1−α

(1− 1

2kK

AL

)αif 0 < K

AL≤ k(

k

2

)αAL if K

AL> k

We have already shown that the gross Marginal Product of Capital is equal to:

MPK = ∂F

∂K=

α

(K

AL

)α−1(1− 1

2kK

AL

)α−1(1− 1

k

K

AL

)if 0 < K

AL≤ k

0 if K

AL> k

In case 0 < K/AL ≤ k, the marginal product of efficiency units of labor ∂F/∂AL is:

∂F

∂AL= (1− α)Kα (AL)−α

(1− 1

2kK

AL

)α+Kα (AL)1−α αK

2kA2L2

(1− 1

2kK

AL

)α−1

= Kα (AL)−α(

1− 12k

K

AL

)α−1 [(1− α)

(1− 1

2kK

AL

)+ α

2kK

AL

]∂F

∂AL=(K

AL

)α(1− 1

2kK

AL

)α−1 [1− α− (1− 2α) 1

2kK

AL

].

The marginal rate of transformation is:

∂F

∂K∂F

∂AL

=α

(K

AL

)α−1(1− 1

2kK

AL

)α−1(1− 1

k

K

AL

)(K

AL

)α(1− 1

2kK

AL

)α−1 [1− α− (1− 2α) 1

2kK

AL

]

=α

(1− 1

k

K

AL

)K

AL

[1− α− (1− 2α) 1

2kK

AL

]∂F

∂K∂F

∂AL

=2α(k − K

AL

)K

AL

[(2− 2α) k − (1− 2α) K

AL

]Expressing everything as a function of:

k ≡ K

AL.

69

we get:∂F/∂K

∂F/∂AL=

2α(k − k

)k[(2− 2α)k − (1− 2α)k

]We can now compute the inverse of the elasticity of substitution:

1σ

= −d log

(∂F/∂K

∂F/∂AL

)d log k

= −(−d log kd log k + d log(k − k)

d log k −d log

[(2− 2α)k − (1− 2α)k

]d log k

)

= −(−1− k

k − k+ (1− 2α)k

(2− 2α)k − (1− 2α)k

)= k

k − k− k

2− 2α1− 2αk − k

1σ

= 1 + k

k − k− k

k − 1− 2α2− 2αk

There, we get the elasticity of substitution as a function of k, which is given by:

σ (k) =

1

1 + k

k − k− k

k − 1− 2α2− 2αk

if 0 < k ≤ k

0 if k ≥ k

For example, with k = 1, and α = 1/3, we arrive at the following elasticity of substitution:

σ (k) =

(4− k) (1− k)(1− k)2 + 3

if 0 < k ≤ k

0 if k ≥ k

E.3 An alternative ModelThe lifetime utility of the representative individual is given by:∫ ∞

0

[γ log

(cTt)

+ (1− γ) log(cNt)]e−βtdt

The flow constraint of the consumer is given by:

bt = rbt + yTt + ptyNt − cTt − ptcNt

Imposing the transversality condition:limt→∞

e−rtbt = 0,

we obtain the intertemporal budget constraint:∫ ∞0

(cTt + ptc

Nt

)e−rtdt = b0 +

∫ ∞0

(yTt + pty

Nt

)e−rtdt.

where yTt and yNt denote the production of tradable and nontradable goods, respectively.

70

Production takes place according to the following technologies:

yTt = ZTt(nTt)α, 0 < α < 1

yNt = ZNt nNt

We may set up the Lagrangian as such:

L =∫ ∞

0u(cTt , c

Nt

)e−βtdt+ λ

[b0 +

∫ ∞0

(yTt + pty

Nt

)e−rtdt−

∫ ∞0

(cTt + ptc

Nt

)e−rtdt

]

We have the following:ucT

(cTt , c

Nt

)= λ

ucN

(cTt , c

Nt

)= λpt

Example. With log utility:γ

cTt= λ

1− γcNt

= λpt

Therefore:cNtcTt

=(

1− γγ

)1pt.

As will become clear below, this condition can be interpreted as a demand function for nontradable goodsrelative to tradable goods. As in standard consumer theory, this demand function depends negatively on therelative price of nontradables, pt.

E.4 A trade-off between unemployment and competitiveness?

71

−16,000Bn

−14,000Bn

−12,000Bn

−10,000Bn

−8,000Bn

−6,000Bn

−4,000Bn

−2,000Bn

0Bn

2,000Bn

80 85 90 95 00 05

Figure 59: Japan Net Exports (1980-2005)

Box 4.2(continued)

2001 crisis.a On average, consumption fell by almost 10 percent in the first year after the crisis. Thisdramatic fall in consumption was accompanied by a shift in the trade balance from an average deficitof 2 percent of GDP to a surplus of 8 percent of GDP and in the current account from a deficit of 5percent to a surplus of 3 percent. In turn the relative price of nontradable goods fell on average byalmost 35 percent.b

Table 4.2 presents three-year averages before and after the nine crises for the three main variables ofinterest. Looking at the individual episodes, we can get a clear sense of how dramatic the turnaroundin the external accounts can be. In Thailand, for example, the shift in the trade balance was closeto 15 percent of GDP and in Malaysia more than 20 percent of GDP. The fall in the relative priceof nontradable may be equally dramatic, even halving in some cases (Argentina, Uruguay, andIndonesia).

In sum, while in practice the shock that triggered the fall in consumption may have differed fromcrisis to crisis, the sharp improvement in the external accounts and the large fall in the relative priceof nontradable goods is fully consistent with the predictions of our simple model.

Table 4.2Crises episodes: Relative price of nontradable goods, trade balance, and current account

Argentina 1982Variable Before Afterp 100 49.6

TB/GDP 0.6% 3.8%

CA/GDP −2.7% −1.9%

Chile 1982Variable Before Afterp 100 70.1

TB/GDP −3.4% 2.1%

CA/GDP −7.3% −7.6%

Uruguay 1982Variable Before Afterp 100 49.2

TB/GDP −4.3% 3.8%

CA/GDP −5.6% −3.0%

Mexico 1994Variable Before Afterp 100 77.9

TB/GDP −4.0% 1.5%

CA/GDP −6.5% −1.0%

Indonesia 1997Variable Before Afterp 100 53.2

TB/GDP 1.6% 14.4%

CA/GDP −2.8% 4.5%

Korea 1997Variable Before Afterp 100 73.7

TB/GDP −1.5% 7.9%

CA/GDP −2.6% −7.1%

Malaysia 1997Variable Before Afterp 100 71.6

TB/GDP 2.6% 25.4%

CA/GDP −6.7% 12.8%

Thailand 1997Variable Before Afterp 100 74.1

TB/GDP −6.9% 7.7%

CA/GDP −5.9% 10.2%

Turkey 2001Variable Before Afterp 100 84.4

TB/GDP −7.8% −4.3%

CA/GDP −1.5% −0.6%

Source: All data from IFS (IMF).Note: “Before” and “After” are three-year averages before and after the crisis. “Before” includes the year ofthe crisis if the crisis took place in July or later. The variable p denotes the relative price of nontradable goods;TB/GDP is the trade balance as a proportion of GDP, and CA/GDP is the current account as a proportion of GDP.

a. We selected these crises because, in all nine of them, consumption fell in the first year after the crisis. The fallin consumption ranges from 5.5 percent in Argentina 1982 to 16 percent in Malaysia 1997. (In Chapter 16, wewill reinterpreted this evidence in light of balance of payments crises models.) The relative price of nontradablegoods was computed as P/EP∗, with P being the domestic CPI, E the nominal exchange rate, and P∗ the US CPI.b. While outside of the scope of the model, it is interesting to notice that consumption recovers relatively quicklyand by the third year after the crisis it has returned to its initial level. (Output behaves similarly.)

Figure 60: Crises episodes: Relative price of nontradable goods, trade balance, and currentaccount (from Vegh (2013))

72

F DataF.1 Definitions of Exchange Rate Regimes

Table 49: Coarse Classification of Exchange Rates (Source: Ilzetzki, Reinhart, and Rogoff(2019))

Coarse Class. Detailed ClassificationFixed / Peg No separate legal tenderFixed / Peg Pre announced peg or currency board arrangementFixed / Peg Pre announced horizontal band that is narrower than or equal to +/-2%Fixed / Peg De facto pegCrawling Peg Pre announced crawling pegCrawling Peg Pre announced crawling band that is narrower than or equal to +/-2%Crawling Peg De factor crawling pegCrawling Peg De facto crawling band that is narrower than or equal to +/-2%Crawling Band Pre announced crawling band that is wider than or equal to +/-2%Crawling Band De facto crawling band that is narrower than or equal to +/-5%Crawling Band Moving band that is narrower than or equal to +/-2% (Crawling Band Managed floating

Floating Freely floating

F.2 Samples

73

Table 50: Summary Statistics (Country Observations)Country Name Sample Period # of obs

Australia 1964-2019 224Austria 1969-2019 204Belgium 1960-2019 240Canada 1960-2019 240

Czech Republic 1993-2019 108Denmark 1969-2019 204Estonia 1989-2019 124

Euro area (17 countries) 1991-2019 116Finland 1960-2019 240France 1960-2019 240

Germany 1992-2019 112Hungary 1992-2019 112Iceland 1964-2019 224Ireland 1990-2019 120Israel 1995-2019 100Italy 1960-2019 240Japan 1960-2019 240Korea 1963-2019 228

Lithuania 2002-2019 72Luxembourg 1985-2019 140

Mexico 1991-2019 116Netherlands 1960-2019 240New Zealand 1960-2019 240

Norway 1972-2019 192OECD - Total 1960-2019 240

Poland 1992-2019 111Portugal 1960-2019 240

Slovak Republic 1993-2019 107South Africa 2000-2019 78

Spain 1976-2019 174Sweden 1960-2019 240

Switzerland 1975-2019 180United Kingdom 1971-2019 196United States 1960-2019 240

NA 1991-2019 116

74

The Phillips Curve: a Relation between Real Exchange Rate ......Phillips Curve Period $ deval...

Documents

Transcript of The Phillips Curve: a Relation between Real Exchange Rate ......Phillips Curve Period $ deval...